# PyTorch Conference North America 2026

- **Event:** PyTorch Conference North America 2026
- **Dates:** Monday, October 19, 2026 - Wednesday, October 21, 2026
- **Location:** San Jose
- **Event website:** https://events.linuxfoundation.org/pytorch-conference-north-america/
- **Timezone:** Pacific Standard Time (UTC-08)
- **Sessions:** 217
- **Speakers:** 280
- **This file:** https://sessionize.com/api/v2/s2rq62nk/view/markdown
- **Session file:** https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/{sessionId}
- **Speaker file:** https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/{speakerId}
- **Generated:** 2026-11-09T08:30:58Z

Follow a session or speaker link for the full details.

## Sessions

### Monday, October 19, 2026

- [Co-Located Event Registration & Badge Pick-Up](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1342264)
  7:30 am - 6:00 pm · The Hub (Street Level)
- [PyTorch, Powering the Enterprise with Red Hat, NVIDIA, and IBM](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1342459)
  9:00 am - 3:30 pm · 210A (Concourse Level)
- [From Academia to Adoption Workshop](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1353046)
  9:00 am - 12:00 pm · 211A
- [LF AI & Data Summit 2026](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1337320)
  9:00 am - 5:00 pm · 210B (Concourse Level)
- [Jupyter Day](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1337325)
  9:00 am - 5:00 pm · 210C (Concourse Level)
- [PyTorch Associate Training](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1353041)
  9:00 am - 5:00 pm · 210D (Concourse Level)
- [Arm Create with Google, Google DeepMind, Nota, and Renesas](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1319956)
  10:00 am - 9:00 pm · The Tech Interactive
- [Fine-Tune LLMs on Amazon SageMaker and Run Them On-Device with GenieX](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1352318)
  1:00 pm - 5:00 pm · Qualcomm Building F (3090 Oakmead Village Dr, Santa Clara, CA 95051)

### Tuesday, October 20, 2026

- [Registration & Badge Pick-Up](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1287931)
  7:30 am - 6:00 pm · The Hub (Street Level)
- [Zen Zone](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1345213)
  7:30 am - 7:35 pm · The Vision Room (Street Level)
- [Keynote: Accelerating Open Intelligence Together](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1298042)
  Mark Collier · 9:00 am - 9:10 am · Grand Ballroom (Concourse Level)
- [Keynote: Shopify's Continual Learning Loop with PyTorch and vLLM](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1298072)
  Cody Mazza-Anthony · 9:10 am - 9:15 am · Grand Ballroom (Concourse Level)
- [Keynote: PyTorch Updates](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1298053)
  Alban Desmaison, Edward Yang · 9:20 am - 9:35 am · Grand Ballroom (Concourse Level)
- [Keynote: MAX and Mojo: The Best Way to Extend PyTorch](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1298067)
  Chris Lattner · 9:35 am - 9:40 am · Grand Ballroom (Concourse Level)
- [Keynote: Ray's Evolution to Support AI Workloads](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1298071)
  Ion Stoica · 9:40 am - 9:45 am · Grand Ballroom (Concourse Level)
- [Keynote: Trainium’s Journey to Native PyTorch](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1298065)
  Robert Hundt · 9:50 am - 9:55 am · Grand Ballroom (Concourse Level)
- [Keynote: Beyond Brute Force: The Era of Adaptive Intelligence.](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1298060)
  Sara Hooker · 9:55 am - 10:03 am · Grand Ballroom (Concourse Level)
- [Keynote: PyTorch at Meta - From Internal Workloads to Multi-Silicon Ecosystem](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1301820)
  Jana van Greunen · 10:05 am - 10:10 am · Grand Ballroom (Concourse Level)
- [Keynote: vLLM Update: Scaling Open Frontier Inference Infrastructure](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1298061)
  Simon Mo · 10:15 am - 10:20 am · Grand Ballroom (Concourse Level)
- [Keynote: Open Source Inference at Production Scale](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1302854)
  Carlos Costa, Abdullah Gharaibeh, Stephen Watt · 10:20 am - 10:27 am · Grand Ballroom (Concourse Level)
- [Keynote: PyTorch Performance for Agents on Vera Rubin](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1298063)
  Ujval Kapasi · 10:30 am - 10:40 am · Grand Ballroom (Concourse Level)
- [Coffee Break](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1287936)
  10:35 am - 11:10 am · Community Expo (Concourse Level)
- [Community Expo](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1287933)
  10:35 am - 7:30 pm · Community Expo (Concourse Level)
- [Contributing to PyTorch with AI agents](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1252250)
  Edward Yang · 10:35 am - 11:05 am · Community Expo (Concourse Level)
- [Meet the Developers of PyTorch](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1364540)
  Alban Desmaison, Natalia Gimelshein, Edward Yang, Driss Guessous, Jane Xu, Will Constable, Tristan Rice · 10:35 am - 11:10 am · Community Expo (Concourse Level)
- [Meet the Developers of Safetensor](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1340758)
  Luc Georges · 10:35 am - 11:10 am · Community Expo (Concourse Level)
- [AWS: Pretrain, Customize, and Serve Foundation Models at Scale: PyTorch on Amazon SageMaker](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1348331)
  10:45 am - 10:55 am · Community Expo (Concourse Level)
- [Demo Theater: Unifying Open-Source LLM Serving on Google Cloud TPUs with TorchTPU](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1304122)
  Rob Mulla · 10:45 am - 10:55 am · Community Expo (Concourse Level)
- [Modal: Computer use agents on Modal](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1361460)
  10:45 am - 10:55 am · Community Expo (Concourse Level)
- [Vultr: Cloud Native Doesn't Mean Cloud Locked.](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1348326)
  10:55 am - 11:05 am · Community Expo (Concourse Level)
- [What's New in Inference in October](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1317774)
  Philip Kiely · 11:10 am - 11:35 am · LL20A (Lower Level)
- [Monarch - Single-Machine DevX at Distributed Scale](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1255147)
  Thomas Wang, Zach DeVito · 11:10 am - 11:35 am · LL20C (Lower Level)
- [How PyTorch Powers Open Foundational Models — From Language to Vision](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256006)
  Songhao Jia, Laura Gustafson, Elisa Cascardi · 11:10 am - 11:35 am · LL21A (Lower Level)
- [Extending TorchInductor with FlyDSL: A New MLIR-Native Backend for High-Performance GEMMs](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256835)
  Liz Li, Cao Jiahui · 11:10 am - 11:35 am · LL21D (Lower Level)
- [Who Owns Production When the Agent Does the Fixing? Governance in AI-Assisted SRE](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1255337)
  Aditi Mewada · 11:10 am - 11:35 am · 210A (Concourse Level)
- [Relay and Reuse: The Dual Engine Behind PyTorch Out-of-Tree Release Readiness](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1253488)
  Zesheng Zong · 11:10 am - 11:35 am · 210B (Concourse Level)
- [A Developer’s Guide to Attention in vLLM](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1254038)
  Lucas Wilkinson, Matthew Bonanni · 11:45 am - 12:10 pm · LL20A (Lower Level)
- [Practical GPU Programming with Triton for PyTorch Developers](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1252848)
  Suman Debnath, JanakiRam Goteti · 11:45 am - 12:10 pm · LL20C (Lower Level)
- [Compute, Latency, and Safety: Architecting Stateful Guardrails for Multi-Agent Workflows](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1251388)
  Purva Chiniya · 11:45 am - 11:55 am · LL21A (Lower Level)
- [Shipping PyTorch and Its Ecosystem: A Modern Release Story](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256844)
  Andrey Talman · 11:45 am - 11:55 am · LL21D (Lower Level)
- [TorchTPU: Running PyTorch Natively on Google TPUs](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1251871)
  Claudio Basile · 11:45 am - 12:10 pm · 210A (Concourse Level)
- [Helion: CuteDSL and TPU Backends for Heterogeneous Hardware, and Why It Suits Agents](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1255964)
  Oguz Ulgen, Dunfan Lu, Jason Ansel · 11:45 am - 12:10 pm · 210B (Concourse Level)
- [Privacy-Preserving AI Across Institutions: Federated Learning Lessons for Medical Imaging](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1199496)
  Sudhanshu Shrivastava · 12:00 pm - 12:10 pm · LL21A (Lower Level)
- [Scaling PyTorch's Compatibility Promise: A Tiered Cross-Repository CI Relay for Out-of-Tree Backends](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1251891)
  Subin George, Jewel K M · 12:00 pm - 12:10 pm · LL21D (Lower Level)
- [Demo Theater: Composable Tensor Programming Across Compiler Stages: PyTorch, TCL, and VISA](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1342322)
  Sol Kim · 12:00 pm - 12:10 pm · Community Expo (Concourse Level)
- [Demo Theater: Your AI Model Needs More Than an API](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1362826)
  Mayank Debnath · 12:15 pm - 12:25 pm · Community Expo (Concourse Level)
- [vLLM Roadmap](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1355211)
  Zachary Xi, Nick Hill · 12:20 pm - 12:45 pm · LL20A (Lower Level)
- [Hardware-Aware AI: Building Agentic Systems from Cloud to Edge with PyTorch, ExecuTorch](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1291518)
  Kavya Sri Chennoju · 12:20 pm - 12:45 pm · LL20C (Lower Level)
- [Closing the Confidence Gap: Making AI Output Trustworthy in Enterprise Systems](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256339)
  Ravi Teja Prabhala Venkata · 12:20 pm - 12:45 pm · LL21A (Lower Level)
- [PyTorch Generalization: A Journey Toward Write Once, Run Anywhere](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256608)
  Yu Guangye, Eikan Wang · 12:20 pm - 12:30 pm · LL21D (Lower Level)
- [The Nemotron Ecosystem: Building the Third Generation — Nano, Super, and Ultra](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256810)
  Shashank Verma, Yi-Fu Wu US · 12:20 pm - 12:45 pm · 210A (Concourse Level)
- [High-Velocity GPU Kernel Authoring with CUTLASS Python](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257904)
  Michael Goldfarb US, Guray Ozen · 12:20 pm - 12:45 pm · 210B (Concourse Level)
- [Demo Theater: From Model to Agent: Running Agentic Coding on Open Infrastructure](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1347549)
  Cam Flanagan · 12:30 pm - 12:40 pm · Community Expo (Concourse Level)
- [Fighting Agents with Agents Bringing Claude to PyTorch CI, triage, and PR review](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1255221)
  Driss Guessous · 12:35 pm - 12:45 pm · LL21D (Lower Level)
- [Attendee Lunch](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1287943)
  12:45 pm - 2:15 pm · Community Expo (Concourse Level)
- [Demo: Fitting Large MoE Fine-Tuning onto Few Nodes: Expert & Context Parallelism in DeepSpeed](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1317873)
  Suman Debnath, Jiale Huang, Daniel Shen · 12:45 pm - 12:55 pm · Community Expo (Concourse Level)
- [Women and Non-Binary in PyTorch Lunch](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1293992)
  12:45 pm - 2:15 pm · 210D (Concourse Level)
- [Demo: Scaling AI Across Devices: ExecuTorch Portability & Performance from Edge to Mobile, XR, & IoT](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1347503)
  Derrick Johnson, Felix Baum · 1:00 pm - 1:10 pm · Community Expo (Concourse Level)
- [Demo Theater: Live Post-Training with TitanRL](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1362598)
  Angel Li · 1:15 pm - 1:25 pm · Community Expo (Concourse Level)
- [Arm: Build a Multimodal Robot Agent with, Reachy Mini and DGX Spark](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1348305)
  1:30 pm - 1:40 pm · Community Expo (Concourse Level)
- [AWS: Native PyTorch on AWS Trainium: Training with PyTorch Ecosystem Libraries](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1348333)
  1:30 pm - 1:40 pm · Community Expo (Concourse Level)
- [Demo Theater: AMD Skills: From Prompt to Performance](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1344446)
  Daniel Holanda · 1:30 pm - 1:40 pm · Community Expo (Concourse Level)
- [Demo Theater: Pretraining at Scale with TorchTitan](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1339821)
  Elfie Guo US · 1:45 pm - 1:55 pm · Community Expo (Concourse Level)
- [NVIDIA: Hands-On with NVIDIA: Demos from the Engineers Behind NVIDIA's AI Stack](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1348317)
  1:50 pm - 2:00 pm · Community Expo (Concourse Level)
- [Demo Theater: IBM Spyre Accelerator in The PyTorch Open Ecosystem](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1351816)
  Mudhakar Srivatsa · 2:00 pm - 2:10 pm · Community Expo (Concourse Level)
- [Smarter Autotuning for Kernels: From Bayesian Optimization to LLM-Guided Search in Helion DSL](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256728)
  Jongsok Choi, Ethan Che · 2:15 pm - 2:40 pm · LL20A (Lower Level)
- [State-of-the-Art KV Transfer for Disaggregated LLM Serving in vLLM](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1255023)
  Nicolò Lucchesi, Sunita Nadampalli, Zhanqiu Hu · 2:15 pm - 2:40 pm · LL20C (Lower Level)
- [Fast, Small, Accurate: Optimizing Models for Inference using FP4 Quantization](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1253987)
  Zhiyu Cheng US, Trenton Starkey · 2:15 pm - 2:40 pm · LL21A (Lower Level)
- [KernelAgent: Hardware-Guided GPU Kernel Optimization via Multi-Agent Orchestration](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1251908)
  Kaiming Cheng, Laura Wang · 2:15 pm - 2:40 pm · LL21D (Lower Level)
- [Demo Theater: Where’s My Training Job?](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1314428)
  Marie Antons · 2:15 pm - 2:25 pm · Community Expo (Concourse Level)
- [From Neurons to Atoms: How PyTorch Powers Open Science at Scale](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1255173)
  Elisa Cascardi, Hubert Banville, Ray Gao · 2:15 pm - 2:40 pm · 210A (Concourse Level)
- [From Scratch to PyTorch: Demystifying ML Frameworks by Building Your Own](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256633)
  Andrea Mattia Garavagno, Vijay Janapa Reddi · 2:15 pm - 2:40 pm · 210B (Concourse Level)
- [Demo Theater: Fighting RL-trained LLMs in Street Fighter III: 3rd Strike](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1304168)
  Andrew Hinh · 2:30 pm - 2:40 pm · Community Expo (Concourse Level)
- [What You Cannot Profile, You Cannot Optimize: Learning to Read PyTorch Traces](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1255448)
  Suvaditya Mukherjee · 2:50 pm - 3:15 pm · LL20A (Lower Level)
- [Native Tiered KV Cache Offloading in vLLM: From Storage Offloading to Disaggregated Serving](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1254938)
  Or Ozeri · 2:50 pm - 3:15 pm · LL20C (Lower Level)
- [SGLang-Diffusion: High Performance Serving Framework for Offline and Real-time Diffusion Models](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1258314)
  Yihao Wang, Kevin Mi · 2:50 pm - 3:15 pm · LL21A (Lower Level)
- [Scaling Foundation Models: From Broken to Near-Linear Multi-Node Training](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256382)
  Sheng Huang · 2:50 pm - 3:00 pm · LL21D (Lower Level)
- [From Fragmented AI to End-to-End Radiology Reports: Scalable Clinical LLMs with TorchTitan](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256115)
  J Deonarine, Khushboo Rathi, Eric Karl · 2:50 pm - 3:15 pm · 210A (Concourse Level)
- [Crossing the Divide: Co-Evolving Kubernetes and Ray for the AI Era](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256376)
  Jago Macleod, Ion Stoica · 2:50 pm - 3:15 pm · 210B (Concourse Level)
- [Training on AWS Trainium via TorchNeuron: TorchTitan and Hugging Face Transformers v5](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256874)
  Maen Suleiman, Michael Benayoun · 3:05 pm - 3:15 pm · LL21D (Lower Level)
- [Understanding Modern Vision Language Models](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1260323)
  Aastha Jhunjhunwala, Mark Moyou · 3:25 pm - 3:50 pm · LL20A (Lower Level)
- [Faster LLM Serving Startup with fastsafetensors](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1255212)
  Takeshi Yoshimura · 3:25 pm - 3:35 pm · LL20C (Lower Level)
- [Making Enterprise Agentic Inference Production-Ready with PyTorch and vLLM](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1301809)
  Joseph Groenenboom, Tyler Michael Smith · 3:25 pm - 3:50 pm · LL21A (Lower Level)
- [EP-Overlap: Hiding Expert-Parallel Communication in MoE Training](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257875)
  Sanket Jayant Purandare, Syed Ahmed · 3:25 pm - 3:35 pm · LL21D (Lower Level)
- [Ray All the Way Down: A Heterogeneous, Elastic PyTorch Training Stack at LinkedIn](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257598)
  Tommy Li, Tao Huang · 3:25 pm - 3:35 pm · 210A (Concourse Level)
- [PerfModel: A Validation-Driven Performance Model for Triton Kernels](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257198)
  Xiaohu Guo · 3:25 pm - 3:50 pm · 210B (Concourse Level)
- [PyTorch Foundation - Accelerator Integration Working Group](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1363752)
  3:30 pm - 4:30 pm · 114 (Street Level)
- [LMCache: a cluster-wide open-source solution for LLM prompt caching](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256217)
  Kuntai Du · 3:40 pm - 3:50 pm · LL20C (Lower Level)
- [Keeping GPUs Busy: High-Speed Storage for PyTorch via fsspec](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1228462)
  Ankita Luthra, Trinadh Kotturu · 3:40 pm - 3:50 pm · LL21D (Lower Level)
- [Gain Geospatial Understanding with the PyTorch Ecosystem](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1258269)
  Steven Pousty · 3:40 pm - 3:50 pm · 210A (Concourse Level)
- [Book Signing with Philip Kiely](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1354949)
  Philip Kiely · 3:50 pm - 4:20 pm · Community Expo (Concourse Level)
- [Coffee Break](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1287937)
  3:50 pm - 4:20 pm · Community Expo (Concourse Level)
- [Guardrails at the Infrastructure Layer: Sandboxing and governing Agentic AI in production](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1244699)
  Sai Charan Teja Gopaluni, Aaresh Sharma · 3:50 pm - 4:20 pm · Community Expo (Concourse Level)
- [Meet the Developers of DeepSpeed](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1364544)
  Minjia Zhang, Zhipeng Wang, Tunji Ruwase, Sam Foreman, Masahiro Tanaka · 3:50 pm - 4:25 pm · Community Expo (Concourse Level)
- [Meet the Developers of Ray](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1340808)
  Mengyin Yan, Richard Liaw, Goutam Venhkatramanan, David Dai, Robert Nishihara · 3:50 pm - 4:25 pm · Community Expo (Concourse Level)
- [Clockwork: Failures should not stop your runs. Watch us kill a GPU or link.](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1348311)
  3:55 pm - 4:05 pm · Community Expo (Concourse Level)
- [Demo Theater: From Prompt to Physical Action: A Live Hardware-Aware AI Demo with PyTorch, ExecuTorch](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1291522)
  Kavya Sri Chennoju · 3:55 pm - 4:05 pm · Community Expo (Concourse Level)
- [Demo Theater: Quantization Showdown: PyTorch Inference Optimization](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1291525)
  Markell Rawls · 4:10 pm - 4:20 pm · Community Expo (Concourse Level)
- [Agentic RL Training in Pytorch](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256335)
  Yichuan Wang, Shuhua Yu · 4:20 pm - 4:45 pm · LL20A (Lower Level)
- [One Model Definition, Many Accelerators: Scaling vLLM Across Hardware Without Forks](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257448)
  Thomas Parnell, Richard Zou · 4:20 pm - 4:45 pm · LL20C (Lower Level)
- [Future of Distributed Communication in PyTorch: New APIs for Fault Tolerance, RDMA and Extensibility](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256220)
  Tristan Rice, Kapil Sharma · 4:20 pm - 4:45 pm · LL21A (Lower Level)
- [JIT Kernel Compilation: How Modular writes fast kernels for any hardware](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257411)
  Stefan Lindall · 4:20 pm - 4:45 pm · LL21D (Lower Level)
- [Beyond Size and Stride: Unleash Performance with Device-Aware Tensor Layouts](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256241)
  Olivier Tardieu, Avery Blanchard · 4:20 pm - 4:30 pm · 210A (Concourse Level)
- [Observability Tooling for Cudagraph Workloads](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1255975)
  Natalia Gimelshein, Driss Guessous · 4:20 pm - 4:30 pm · 210B (Concourse Level)
- [Nested Graph Breaks: Reducing the Cost of Graph Breaks in torch.compile](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256035)
  William Wen · 4:35 pm - 4:45 pm · 210A (Concourse Level)
- [rocSHMEM Symmetric Memory in PyTorch for AMD GPUs](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256215)
  Prachi Gupta · 4:35 pm - 4:45 pm · 210B (Concourse Level)
- [SkyRL: Democratizing Scalable RL Training](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1345867)
  Sumanth Hegde, Eric Tang · 4:55 pm - 5:20 pm · LL20A (Lower Level)
- [Portable PyTorch Across AI Accelerators: A Triton Operator Stack from Eager Mode to vLLM](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256909)
  Yonghua Lin · 4:55 pm - 5:20 pm · LL20C (Lower Level)
- [Efficient MoE LLM Inference on Arm with vLLM and OpenVINO](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257004)
  Abhishek Jain, N Maajid Khan · 4:55 pm - 5:05 pm · LL21A (Lower Level)
- [Integrating the IBM Spyre Accelerator](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256168)
  David Grove, Antoni Viros i Martin, Avery Blanchard · 4:55 pm - 5:20 pm · LL21D (Lower Level)
- [Five Bottlenecks, Five Fixes: Systematic Performance Diagnosis for PyTorch Distributed Training](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1254803)
  Paulo Aragao, Aravind Neelakantan · 4:55 pm - 5:20 pm · 210A (Concourse Level)
- [Unlocking PyTorch for Dataflow Accelerators: An Open-Source Kernel Tile IR and Dataflow Scheduler](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257807)
  Prasanth Chatarasi, Viji Srinivasan, Swagath Venkataramani · 4:55 pm - 5:20 pm · 210B (Concourse Level)
- [XCCL: Scaling PyTorch Collectives to Exascale on Intel GPUs with TorchComms](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1258012)
  Panagiotis Kourdis, Tanima Dey · 5:10 pm - 5:20 pm · LL21A (Lower Level)
- [Static Tensor Shape Checking for PyTorch with Pyrefly](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256337)
  Steven Troxler, Avik Chaudhuri · 5:30 pm - 5:55 pm · LL20A (Lower Level)
- [vLLM KV Cache Management for Model-Specific Requirements](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1253418)
  Xiyuan Wang · 5:30 pm - 5:40 pm · LL20C (Lower Level)
- [TorchNeuron: Native PyTorch RL on AWS Trainium - From Research to Production Without Compromise](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1248603)
  Yahav Biran, Alex Nails · 5:30 pm - 5:40 pm · LL21A (Lower Level)
- [Open Source Reinforcement Learning with Agent Harnesses](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1253573)
  Ben Burtenshaw · 5:30 pm - 5:55 pm · LL21D (Lower Level)
- [PyTorch for Agentic AI: Scaling Heterogeneous Systems from CPU to XPU](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1291520)
  Eikan Wang, Huma Abidi · 5:30 pm - 5:55 pm · 210A (Concourse Level)
- [From Weeks to Overnight: Autonomous Day-0 Kernel Bring-Up with Agent Pipelines](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256713)
  Juan Muneton Gallego · 5:30 pm - 5:40 pm · 210B (Concourse Level)
- [Building Portable, Composable Local Agents with ExecuTorch](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257786)
  Mergen Nachin, Digant Desai · 5:45 pm - 5:55 pm · LL20C (Lower Level)
- [Dist-MoE: Fused Comm-Compute Kernels & Activation Packing for Speed-of-Light MoEs](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257887)
  Sanket Jayant Purandare, Shikai Li · 5:45 pm - 5:55 pm · LL21A (Lower Level)
- [Before the Loss Spike: Training Alignment for Precise PyTorch Debugging at Scale](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1199488)
  Ziming Zhou · 5:45 pm - 5:55 pm · 210B (Concourse Level)
- [Flare Party & Poster Presentations](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1287945)
  6:00 pm - 7:30 pm · Community Expo (Concourse Level)
- [Poster Presentations](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1311764)
  6:00 pm - 7:30 pm · Community Expo (Concourse Level)

### Wednesday, October 21, 2026

- [Registration & Badge Pick-Up](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1287932)
  8:00 am - 7:00 pm · The Hub (Street Level)
- [Zen Zone](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1345283)
  8:00 am - 8:35 pm · The Vision Room (Street Level)
- [Keynote: Building What’s Next in Open Source AI](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1303522)
  Mark Collier · 9:00 am - 9:10 am · Grand Ballroom (Concourse Level)
- [Keynote: From AI to Agentic AI and Back](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1298047)
  Mazin Gilbert · 9:10 am - 9:15 am · Grand Ballroom (Concourse Level)
- [Keynote: Workload Fungibility in the Agentic Era: Bringing Native PyTorch to Cloud TPUs](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1298062)
  Bill Jia, Zongwei Zhou · 9:20 am - 9:30 am · Grand Ballroom (Concourse Level)
- [Keynote: From Silicon to Models at Scale](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1298073)
  Priya Nagpurkar · 9:30 am - 9:35 am · Grand Ballroom (Concourse Level)
- [Keynote: Contributor Awards - Sponsored by Red Hat](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1303527)
  9:40 am - 9:55 am · Grand Ballroom (Concourse Level)
- [Keynote: Scaling Large Frontier Models Training with DeepSpeed](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1314372)
  Zhipeng Wang · 9:55 am - 9:58 am · Grand Ballroom (Concourse Level)
- [Keynote: Open Science and AI Foundation Models: Allies in the Pursuit of Knowledge](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1298057)
  Rachad Alao · 10:00 am - 10:08 am · Grand Ballroom (Concourse Level)
- [Keynote: AI Model Customization Platform](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1298069)
  Yiwei Song · 10:10 am - 10:15 am · Grand Ballroom (Concourse Level)
- [Keynote: Linear Algebra for the Age of Research](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1298055)
  Mark Saroufim · 10:15 am - 10:23 am · Grand Ballroom (Concourse Level)
- [Coffee Break](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1287938)
  10:35 am - 11:10 am · Community Expo (Concourse Level)
- [Community Expo](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1287934)
  10:35 am - 4:20 pm · Community Expo (Concourse Level)
- [Contributing to Inference OSS That Won't Stand Still: A BoF on vLLM, llm-d, and the Moving Target](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257610)
  Maroon Ayoub, Nili Guy · 10:35 am - 11:05 am · Community Expo (Concourse Level)
- [Meet the Developers of vLLM](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1340810)
  George Novack, Nick Hill · 10:35 am - 11:10 am · Community Expo (Concourse Level)
- [Demo: PyTorch Ecosystem Running Natively on Trainium: Reinforcement Learning - TorchTitan & Monarch](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1291526)
  Maen Suleiman · 10:40 am - 10:50 am · Community Expo (Concourse Level)
- [Demo Theater: The Engineering Behind 3-Second Video Generation](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1314406)
  Ali Taha · 10:55 am - 11:05 am · Community Expo (Concourse Level)
- [NVIDIA: Hands-On with NVIDIA: Demos from the Engineers Behind NVIDIA's AI Stack](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1348321)
  10:55 am - 11:05 am · Community Expo (Concourse Level)
- [Red Hat: Quantization Showdown: PyTorch Inference Optimization on OpenShift AI](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1348510)
  10:55 am - 11:05 am · Community Expo (Concourse Level)
- [Beyond the Brrr: Building a Unified Ecosystem for Optimized Kernels](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1252682)
  Ben Burtenshaw · 11:10 am - 11:35 am · LL20A (Lower Level)
- [FlexShard: Pluggable Sharding Placements for Matrix Optimizers and Block-wise Quantization](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256015)
  Wei Feng, Anshul Sinha, Ailing Zhang · 11:10 am - 11:35 am · LL20C (Lower Level)
- [TPU agent suite: Accelerating PyTorch Model Onboarding and Inference on Cloud TPUs](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1301802)
  Sandeep Pokkunuri, Chris Jones · 11:10 am - 11:35 am · LL21A (Lower Level)
- [DeepSpeed Is Not Just ZeRO: Tensor, Sequence, and Expert Parallelism with ZeRO-Level Usability](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257990)
  Masahiro Tanaka · 11:10 am - 11:35 am · LL21D (Lower Level)
- [Inside the PyTorch Ambassador Model: How Ambassadors Grow PyTorch Adoption & How You Can Become One](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1306148)
  Sahdev Zala, Sonny Mupfuni · 11:10 am - 11:35 am · 210A (Concourse Level)
- [Building Highly Efficient Inference System for Recommenders Using PyTorch](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1243435)
  Lu Fang, Ilina Mitra · 11:10 am - 11:35 am · 210B (Concourse Level)
- [Deploying PyTorch Models Where Failure Isn't an Option](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1251868)
  Sarah Mohamed, Ashvant Ram Selvam · 11:45 am - 12:10 pm · LL20A (Lower Level)
- [Transformers.distributed: Bringing First-Class Distributed Training to 2.4M+ Hugging Face Models](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1253220)
  Ferdinand Mom · 11:45 am - 11:55 am · LL20C (Lower Level)
- [Parametrized Dynamic Shape CUDA Graphs](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257948)
  Elias Ellison, Daniel Galvez · 11:45 am - 12:10 pm · LL21A (Lower Level)
- [Miles: Enterprise-facing Agentic RL Framework](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1258156)
  Yueming Yuan · 11:45 am - 12:10 pm · LL21D (Lower Level)
- [What Would a PyTorch Expert Say? Turning Tribal AI Knowledge into a Model](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1333197)
  Liana Koleva, Teja Pulagam · 11:45 am - 12:10 pm · 210A (Concourse Level)
- [Beyond Collectives: Building Fused Compute-Communication Kernels with PyTorch Symmetric Memory](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257030)
  Ke Wen, Natalia Gimelshein, Kapil Sharma · 11:45 am - 12:10 pm · 210B (Concourse Level)
- [Precompile for Training](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1245005)
  Bob Ren, Aaron Orenstein · 12:00 pm - 12:10 pm · LL20C (Lower Level)
- [Scaling Audio AI Infrastructure: Building Voice-Native AI](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1258143)
  Mu Li, Huapeng Zhou, Lindsey Allen · 12:20 pm - 12:45 pm · LL20A (Lower Level)
- [Self-Healing PyTorch Training Through Live GPU Migration](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1322707)
  Pavneet Ahluwalia, Gavin Cohen · 12:20 pm - 12:45 pm · LL20C (Lower Level)
- [From torch.profiler to Hardware Cycles: A Practical Profiling Playbook for AWS Trainium](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257690)
  Esha Lakhotia, Pinak Panigrahi · 12:20 pm - 12:45 pm · LL21A (Lower Level)
- [Train the Agent, Not Just the Model](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1255556)
  Sergio Paniego Blanco · 12:20 pm - 12:45 pm · LL21D (Lower Level)
- [Demo Theater: WideEP NVL72 with Rust Frontend](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1361627)
  Rain Jiang, Kevin Li · 12:20 pm - 12:30 pm · Community Expo (Concourse Level)
- [Ecosystem Working Group: Unlocking Community Impact Through the PyTorch Landscape](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1253136)
  Joseph Groenenboom, Chris Hoge · 12:20 pm - 12:45 pm · 210A (Concourse Level)
- [Making vLLM Faster on Intel GPUs with Triton Kernels](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1255903)
  Whitney Tsang, Artur Fierka · 12:20 pm - 12:30 pm · 210B (Concourse Level)
- [Demo Theater: dmx-compressor: Accelerating the Development of Kernels for Custom ASIC Hardware](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1309107)
  Tristan Webb · 12:35 pm - 12:45 pm · Community Expo (Concourse Level)
- [FlexGEMM: Flexible PyTorch Epilogues](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1252009)
  Driss Guessous · 12:35 pm - 12:45 pm · 210B (Concourse Level)
- [Attendee Lunch](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1287944)
  12:45 pm - 2:15 pm · Community Expo (Concourse Level)
- [Demo Theater: Never Hit Zero: A Live Demo of Surviving Link and Node Failures in PyTorch Training](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1322762)
  Lin She · 12:45 pm - 12:55 pm · Community Expo (Concourse Level)
- [Demo Theater: AI Playground Home Agent: Remote PyTorch AI Workflows from Your Phone](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1291528)
  Ashok Emani, Qiacheng Li · 1:00 pm - 1:10 pm · Community Expo (Concourse Level)
- [Demo Theater: Agile Readiness: Leveraging CI Relay for PyTorch Workflows](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1361630)
  Zesheng Zong · 1:15 pm - 1:25 pm · Community Expo (Concourse Level)
- [Demo Theater: TPU Megakernels](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1360619)
  George Novack · 1:45 pm - 1:55 pm · Community Expo (Concourse Level)
- [Demo Theater: Why is Heterogeneous Computing So Hard and Why Does it Have To Be?](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1291533)
  Jay Dawani · 2:00 pm - 2:10 pm · Community Expo (Concourse Level)
- [MXFP8/4 pretraining at scale with Torchtitan on AMD Instinct GPUs](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256822)
  Liz Li, Shekhar Pandey · 2:15 pm - 2:40 pm · LL20A (Lower Level)
- [PyTorch-Native LLM Serving on TPU: SGLang and vLLM](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256716)
  Colin Taylor, Qi Zhou, Angela Yi · 2:15 pm - 2:40 pm · LL20C (Lower Level)
- [From Pixels to Physical Motion: Building World Action Models (WAM) with PyTorch & NVIDIA Cosmos](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257041)
  Susie Xia, Ruijie Zheng US, George Kurian · 2:15 pm - 2:40 pm · LL21A (Lower Level)
- [Advancing torch.compile for Verifiable Precision & Dynamic Shapes](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1253426)
  Xiyuan Wang · 2:15 pm - 2:40 pm · LL21D (Lower Level)
- [Demo Theater: Beyond torch.compile: Reducing Data Movement with Device-Persistent Tensors in PyTorch](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1291535)
  Minwook Ahn · 2:15 pm - 2:25 pm · Community Expo (Concourse Level)
- [Clearing the Path Towards an ABI Stable PyTorch C++ Extension Ecosystem](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1252992)
  Sean McGovern, Chris Leonard, Jane Xu · 2:15 pm - 2:40 pm · 210A (Concourse Level)
- [From Full DTensor to spmd\_types: Making Parallelism Explicit in TorchTitan](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257076)
  Chien-Chin Huang, Pian Pawakapan · 2:15 pm - 2:40 pm · 210B (Concourse Level)
- [Demo Theater: OpenCLIC: Open Communication for Distributed Inference](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1329720)
  Jesse Treger · 2:30 pm - 2:40 pm · Community Expo (Concourse Level)
- [HiFloat: Democratizing Ultra-Low Precision Training and Inference in PyTorch Ecosystems](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1206584)
  Yun Zhao, Haonan Zhang · 2:50 pm - 3:00 pm · LL20A (Lower Level)
- [Elastic Expert Parallelism in vLLM](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1252137)
  Itay Alroy · 2:50 pm - 3:15 pm · LL20C (Lower Level)
- [Native DSL Operators in PyTorch Core](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1255976)
  Simon Layton · 2:50 pm - 3:15 pm · LL21A (Lower Level)
- [Beyond AdamW: A Practical PyTorch Walkthrough of Muon, Dion, and Orthogonalized Optimizer Variants](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1248446)
  Jennifer Wei · 2:50 pm - 3:15 pm · LL21D (Lower Level)
- [From PyTorch to Production: Serving a Physics-Constrained Generative Model with ONNX, Ray, and vLLM](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256403)
  Arun Sharma · 2:50 pm - 3:15 pm · 210A (Concourse Level)
- [Agentic Inference on Rubin: Challenges and Optimizations](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1337217)
  Alex Settle · 2:50 pm - 3:15 pm · 210B (Concourse Level)
- [TorchJD: Training PyTorch Models with Multiple Losses](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256805)
  Valérian Rey, Khush Patel · 3:05 pm - 3:15 pm · LL20A (Lower Level)
- [Efficient Pretraining of LLMs in NVFP4](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256332)
  Anjulie Agrusa US, Ryan Spring US, Bruce Zitelli · 3:25 pm - 3:50 pm · LL20A (Lower Level)
- [Building a Post-Training Platform for Teams That Don't Own the Training Loop](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257963)
  Gaurav Arora, Shunyao Li, Eric Wang · 3:25 pm - 3:50 pm · LL20C (Lower Level)
- [Boosting PyTorch’s Heterogeneous Compute Ecosystem: Optimizing LLM Training and Serving on NPU](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1348281)
  Fred Li · 3:25 pm - 3:50 pm · LL21A (Lower Level)
- [TorchTitan GraphTrainer: LLM Distributed Training with PyTorch's Compiler Toolkit](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257898)
  Sanket Jayant Purandare, Sherlock Huang · 3:25 pm - 3:50 pm · LL21D (Lower Level)
- [Deploying Robot Policies Across Many Targets with ExecuTorch](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257831)
  Anthony Shoumikhin, Jacob Szwejbka · 3:25 pm - 3:50 pm · 210A (Concourse Level)
- [Portable Paged Attention: From Triton to Helion](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1255851)
  Burkhard Ringlein · 3:25 pm - 3:50 pm · 210B (Concourse Level)
- [“What Do I Do Now?” Bridging the ExecuTorch Adoption Gap Through Developer Learning](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1255970)
  Matt Cossins · 3:50 pm - 4:20 pm · Community Expo (Concourse Level)
- [Coffee Break](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1287940)
  3:50 pm - 4:20 pm · Community Expo (Concourse Level)
- [Meet the Developers of Helion](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1340815)
  Jason ANSEL, Jongsok CHOI, Yarong Mu, Freya Azad, Theotime Combes, Ethan Che, Dunfan Lu, Shangdi Yu, Oguz Ulgen, Emilio Cota, Karthick Panner Selvam · 3:50 pm - 4:25 pm · Community Expo (Concourse Level)
- [Demo Theater: Evolving Ray Core for Post-Training at Scale](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1345871)
  Mengjin Yan, Joshua Lee · 3:55 pm - 4:05 pm · Community Expo (Concourse Level)
- [Prefix Caching for Autoregressive Stages in Multi-Stage Pipelines](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1255511)
  Ricardo Noriega, Alex Brooks · 4:20 pm - 4:45 pm · LL20A (Lower Level)
- [Primus Tuning: Hybrid Projection and Agentic Search for Distributed Training](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256459)
  Anshu Raina, Peyman Razaghi · 4:20 pm - 4:45 pm · LL20C (Lower Level)
- [From Bytes to Tensors: Efficient Media Processing with PyTorch in 2026](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1254966)
  Nicolas Hug, Scott Schneider · 4:20 pm - 4:45 pm · LL21A (Lower Level)
- [Torchtitan RL: A Unified and Extensible Training Framework for Agentic Tasks](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256870)
  Felipe Mello, Jiani Wang · 4:20 pm - 4:45 pm · LL21D (Lower Level)
- [NVFP4 Training Tips and Tricks](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1347085)
  Paul Chang, Erik Schultheis · 4:20 pm - 4:45 pm · 210A (Concourse Level)
- [From Backed to Unbacked: Sound, Predictable, and Controllable Dynamic Shapes in PyTorch](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1260092)
  Laith Sakka · 4:20 pm - 4:45 pm · 210B (Concourse Level)
- [Building Cost-Efficient Mixture-of-Models on AMD GPUs with vLLM Semantic Router](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1333195)
  Mahdi Ghodsi · 4:55 pm - 5:20 pm · LL20A (Lower Level)
- [PyTorch-Native Feature Transformation and Training Framework for Uber Eats Recommendation](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1241540)
  Peng Zhang, Ke Chen, Xandra Zhu · 4:55 pm - 5:20 pm · LL20C (Lower Level)
- [Speeding Up torch.compile: A New FakeTensor](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257046)
  Angel Li · 4:55 pm - 5:05 pm · LL21A (Lower Level)
- [When Rollout and Training Disagree: Lessons from Building a Mismatch-Free RL Engine](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257767)
  Neiwen Ling, Tianle Zhong · 4:55 pm - 5:05 pm · LL21D (Lower Level)
- [Introducing NCCL Extensions: Communication Patterns for Modern AI](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257023)
  Sreeram Potluri, Artem Polyakov · 4:55 pm - 5:20 pm · 210A (Concourse Level)
- [Scaling PyTorch Dataloaders 10x: A Case Study](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1243544)
  Ian Stenbit, Christine Cheng, Dylan Doblar · 4:55 pm - 5:20 pm · 210B (Concourse Level)
- [Lightweight FX Tracing in PyTorch](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256183)
  Yidi Wu, Richard Zou · 5:10 pm - 5:20 pm · LL21A (Lower Level)
- [Integrating TorchTPU into TorchTitan and Optimizing Models for TPUs](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1246272)
  Aleksey Vlasenko, Will Constable · 5:10 pm - 5:20 pm · LL21D (Lower Level)
- [From PyTorch to the Edge: Agentic Synthesis of Inference Runtimes for Heterogeneous Hardware](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257310)
  Thomas Cottenier · 5:30 pm - 5:55 pm · LL20A (Lower Level)
- [Scaling PyTorch on Intel XPU: Field Notes from AI for Science on Aurora](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257912)
  Sam Foreman, Panagiotis Kourdis, Tanima Dey · 5:30 pm - 5:55 pm · LL20C (Lower Level)
- [Unlocking the Full Potential of TorchDynamo: Accelerating, Comparing, and Debugging ML Systems](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257418)
  Yi Pan, Stephanie Wang · 5:30 pm - 5:55 pm · LL21A (Lower Level)
- [Lessons at Scale: Building a PyTorch Inference Stack with Open-Source Components](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1347541)
  Oscar Savolainen · 5:30 pm - 5:55 pm · LL21D (Lower Level)
- [MCCL: Fault-Tolerant Collective Communication for Large-Scale GPU Training](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256858)
  Ben Carver · 5:30 pm - 5:55 pm · 210A (Concourse Level)
- [Multimodal Dataloaders for Physical AI](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257585)
  Peter Prettenhofer · 5:30 pm - 5:55 pm · 210B (Concourse Level)
- [AI Community Bash featuring DE LA SOUL](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1311946)
  DE LA SOUL · 6:00 pm - 8:30 pm · Concert Venue (Concourse Level)

## Speakers

- [Aaresh Sharma](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/004a5326-eb16-403f-98ba-b15cf018b27c) — Sr. Specialist SA - Containers
  [Guardrails at the Infrastructure Layer: Sandboxing and governing Agentic AI in production](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1244699)
- [Aaron Orenstein](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/750d7a61-d753-4b28-b74d-dcb57d323772) — Software Engineer, Meta
  [Precompile for Training](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1245005)
- [Aastha Jhunjhunwala](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/85704f37-4b3d-47ef-9a67-f5cd637f6f9f) — Solutions Architect
  [Understanding Modern Vision Language Models](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1260323)
- [Abdullah Gharaibeh](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/44e2d18c-34f4-416b-8c7c-3240cf253182)
  [Keynote: Open Source Inference at Production Scale](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1302854)
- [Abhishek Jain](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/a5a6e682-46e0-4a33-9aa6-38e2bc6a4543) — Software Engineering Manager, Fujitsu Research
  [Efficient MoE LLM Inference on Arm with vLLM and OpenVINO](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257004)
- [Aditi Mewada](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/fe7da75f-1d59-4c92-9bf8-b84854b6da3e) — Software Engineer - Apple
  [Who Owns Production When the Agent Does the Fixing? Governance in AI-Assisted SRE](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1255337)
- [Ailing Zhang](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/226fe4a3-3622-4330-b399-eb6473ffb992) — Software Engineer, PyTorch at Meta
  [FlexShard: Pluggable Sharding Placements for Matrix Optimizers and Block-wise Quantization](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256015)
- [Alban Desmaison](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/31e70d6a-bdb7-4be2-9f3e-0407ebf617e0) — Research Engineer at Meta
  [Keynote: PyTorch Updates](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1298053), [Meet the Developers of PyTorch](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1364540)
- [Aleksey Vlasenko](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/73627a13-13b8-4d01-b0ea-82b706eeb706) — Google, SWE
  [Integrating TorchTPU into TorchTitan and Optimizing Models for TPUs](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1246272)
- [Alex Brooks](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/94399a2d-83f2-431d-9354-7715e3b7e4af) — Principal Machine Learning Engineer @ Red Hat
  [Prefix Caching for Autoregressive Stages in Multi-Stage Pipelines](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1255511)
- [Alex Nails](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/26ee5432-b90a-4ea7-b49b-9e54cb00fab9) — MTS, RadixArk. Core Maintainer of SGLang
  [TorchNeuron: Native PyTorch RL on AWS Trainium - From Research to Production Without Compromise](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1248603)
- [Alex Settle](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/09164dfb-87c7-41e9-8799-5f3339faa422)
  [Agentic Inference on Rubin: Challenges and Optimizations](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1337217)
- [Ali Taha](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/0af5cb00-6ea0-45fc-bc13-69e9a76a4290) — Baseten’s Waterloo Intern
  [Demo Theater: The Engineering Behind 3-Second Video Generation](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1314406)
- [Andrea Mattia Garavagno](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/796e9a90-c420-435c-adcd-c89ea3da2793) — PostDoc at ETH Zurich | Author, Introduction to Microprocessor-Based Systems Design
  [From Scratch to PyTorch: Demystifying ML Frameworks by Building Your Own](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256633)
- [Andrew Hinh](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/f94de2de-8ad5-4b14-8caa-45f27535b3bc) — Developer Relations Engineer @ Modal
  [Demo Theater: Fighting RL-trained LLMs in Street Fighter III: 3rd Strike](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1304168)
- [Andrey Talman](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/43779dfc-5158-48d0-9dad-1a82d5e1c959) — PyTorch Dev Infra, Software Engineer
  [Shipping PyTorch and Its Ecosystem: A Modern Release Story](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256844)
- [Angel Li](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/548bfed7-9240-43da-85bf-b62b5bef6117) — Software Engineer, Meta
  [Demo Theater: Live Post-Training with TitanRL](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1362598), [Speeding Up torch.compile: A New FakeTensor](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257046)
- [Angela Yi](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/a67d0c53-0f06-4c9b-9b77-c6524ca443c5) — Meta, PyTorch Compiler, Software Engineer
  [PyTorch-Native LLM Serving on TPU: SGLang and vLLM](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256716)
- [Anjulie Agrusa US](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/7f3f66c1-edd4-4cd4-9c6e-729c3e2937df) — NVIDIA Senior Machine Learning Engineer, Quantization for Low Precision LLM Training
  [Efficient Pretraining of LLMs in NVFP4](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256332)
- [Ankita Luthra](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/41892054-a6c5-41c3-832f-10ee323d6417) — Senior Software Engineer, Google Cloud Storage
  [Keeping GPUs Busy: High-Speed Storage for PyTorch via fsspec](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1228462)
- [Anshu Raina](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/e0a5dd7e-1f74-4388-be77-70ee27137f95) — AMD
  [Primus Tuning: Hybrid Projection and Agentic Search for Distributed Training](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256459)
- [Anshul Sinha](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/e8243db5-a20f-4da2-98e7-e94fd3c23ccd) — Software Engineer @ Meta
  [FlexShard: Pluggable Sharding Placements for Matrix Optimizers and Block-wise Quantization](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256015)
- [Anthony Shoumikhin](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/efb12805-830f-47e0-a33a-cc2c5cb30515) — Meta
  [Deploying Robot Policies Across Many Targets with ExecuTorch](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257831)
- [Antoni Viros i Martin](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/b40cf3ab-e12a-4968-ae89-8abfc4199707) — Staff Research Scientist, IBM Research
  [Integrating the IBM Spyre Accelerator](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256168)
- [Aravind Neelakantan](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/5ebe0f49-ab78-42b4-85c7-9def6615cae7) — Specialist Solutions Architect, Amazon Web Services
  [Five Bottlenecks, Five Fixes: Systematic Performance Diagnosis for PyTorch Distributed Training](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1254803)
- [Artem Polyakov](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/38de4619-189d-4cc3-bb0d-0ef3166f0637) — Sr. SW Engineer at NVIDIA
  [Introducing NCCL Extensions: Communication Patterns for Modern AI](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257023)
- [Artur Fierka](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/570164ec-90eb-4665-a47c-6dbbef75f2fd) — AI Software Solutions Engineer @ Intel
  [Making vLLM Faster on Intel GPUs with Triton Kernels](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1255903)
- [Arun Sharma](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/a68d7c19-6355-486a-ac5a-7d1548ba7c14) — PhD Researcher, Spatial AI and Physics-Informed ML, University of Minnesota
  [From PyTorch to Production: Serving a Physics-Constrained Generative Model with ONNX, Ray, and vLLM](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256403)
- [Ashok Emani](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/5cd975bf-b1b5-4353-94cf-d1a275f251c7) — AI Frameworks Engineer, Intel
  [Demo Theater: AI Playground Home Agent: Remote PyTorch AI Workflows from Your Phone](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1291528)
- [Ashvant Ram Selvam](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/6a8ecd1f-16a6-463c-8be9-4ec62015f0e8) — Software Engineer, MathWorks
  [Deploying PyTorch Models Where Failure Isn't an Option](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1251868)
- [Avery Blanchard](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/a5640c88-ada2-48d1-bba1-29c885a6bd4e) — IBM
  [Beyond Size and Stride: Unleash Performance with Device-Aware Tensor Layouts](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256241), [Integrating the IBM Spyre Accelerator](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256168)
- [Avik Chaudhuri](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/a622f099-a1fa-48c5-8786-e29a80af0710) — Software Engineer, PyTorch Compiler @ Meta
  [Static Tensor Shape Checking for PyTorch with Pyrefly](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256337)
- [Ben Burtenshaw](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/4940e018-46ea-427f-aca0-09929b255e2d) — Machine Learning, Hugging Face
  [Open Source Reinforcement Learning with Agent Harnesses](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1253573), [Beyond the Brrr: Building a Unified Ecosystem for Optimized Kernels](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1252682)
- [Ben Carver](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/92f2fbc0-20d1-49af-9474-c8055566f5bd) — Research Scientist, Meta Platforms
  [MCCL: Fault-Tolerant Collective Communication for Large-Scale GPU Training](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256858)
- [Bill Jia](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/d1da9d84-4e73-4d4b-b5ca-40ea7cdf49b3)
  [Keynote: Workload Fungibility in the Agentic Era: Bringing Native PyTorch to Cloud TPUs](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1298062)
- [Bob Ren](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/fe5f994f-4af5-4ec0-b379-d17a581b641b) — Software Engineer
  [Precompile for Training](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1245005)
- [Bruce Zitelli](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/62818039-d592-4267-af73-b2d1867469ae) — Software Engineer
  [Efficient Pretraining of LLMs in NVFP4](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256332)
- [Burkhard Ringlein](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/46087d01-33fe-40f2-8ea1-8c63ab43d38c) — Research Staff Member at IBM Research
  [Portable Paged Attention: From Triton to Helion](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1255851)
- [Cam Flanagan](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/5e6ca8c5-f679-4535-85ea-8c708138c8b4) — Senior AI Product Engineer, Nscale
  [Demo Theater: From Model to Agent: Running Agentic Coding on Open Infrastructure](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1347549)
- [Cao Jiahui](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/8043cfad-15fa-4cf5-9c9f-b3eea668888a) — Sr. Software Development Engineer@AMD
  [Extending TorchInductor with FlyDSL: A New MLIR-Native Backend for High-Performance GEMMs](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256835)
- [Carlos Costa](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/84e0bdc8-3b4c-49cb-9cbe-7cdb4e9092b9)
  [Keynote: Open Source Inference at Production Scale](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1302854)
- [Chien-Chin Huang](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/082f4b54-aa62-49ee-9fbc-b21bca778c8e) — Software Engineer, PyTorch Distributed
  [From Full DTensor to spmd\_types: Making Parallelism Explicit in TorchTitan](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257076)
- [Chris Hoge](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/52aea18c-5412-4ed3-bce9-45eb96ca56f1) — AI Platform Software at NVIDIA
  [Ecosystem Working Group: Unlocking Community Impact Through the PyTorch Landscape](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1253136)
- [Chris Jones](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/79deec14-44bd-4083-9390-e74e169280fa) — Google, Senior Product Manager
  [TPU agent suite: Accelerating PyTorch Model Onboarding and Inference on Cloud TPUs](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1301802)
- [Chris Lattner](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/19154a93-c5a7-4c08-9466-378ff9857c47)
  [Keynote: MAX and Mojo: The Best Way to Extend PyTorch](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1298067)
- [Chris Leonard](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/b200528e-d1d7-4b96-8f94-7dc4b2d96c78) — Red Hat - Senior Machine Learning Engineer
  [Clearing the Path Towards an ABI Stable PyTorch C++ Extension Ecosystem](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1252992)
- [Christine Cheng](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/cddb178c-1c7c-466c-9d46-e71ecd674fe5) — Software Engineer, Nvidia
  [Scaling PyTorch Dataloaders 10x: A Case Study](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1243544)
- [Claudio Basile](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/d2a8722c-ec18-4a85-9558-309c5358044e) — Tech Lead - TorchTPU @ Google
  [TorchTPU: Running PyTorch Natively on Google TPUs](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1251871)
- [Cody Mazza-Anthony](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/a7425c62-8a97-4b05-b7ec-9c3c3e1b2461)
  [Keynote: Shopify's Continual Learning Loop with PyTorch and vLLM](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1298072)
- [Colin Taylor](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/0f935305-94c0-4bf8-a921-eff93b3b6789) — Software Engineer, Pytorch
  [PyTorch-Native LLM Serving on TPU: SGLang and vLLM](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256716)
- [Daniel Galvez](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/532fdd19-365a-4f9f-be3e-d6e0e3ba91a1) — NVIDIA, AI Developer Technology Engineer
  [Parametrized Dynamic Shape CUDA Graphs](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257948)
- [Daniel Holanda](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/79fe72f0-3e43-40a5-93f3-db5f61f66981) — Principal AI Engineer
  [Demo Theater: AMD Skills: From Prompt to Performance](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1344446)
- [Daniel Shen](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/4738ab3a-ecd6-4b22-9784-90167213cf15) — Machine Learning Engineer
  [Demo: Fitting Large MoE Fine-Tuning onto Few Nodes: Expert & Context Parallelism in DeepSpeed](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1317873)
- [David Dai](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/586de102-1ddf-452e-ac08-adf0e0eb5d09)
  [Meet the Developers of Ray](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1340808)
- [David Grove](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/5f649438-ac92-4bbc-bbba-ec613816ae96) — IBM
  [Integrating the IBM Spyre Accelerator](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256168)
- [DE LA SOUL](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/850357c2-3885-4621-8e37-b7ede68f9fe2)
  [AI Community Bash featuring DE LA SOUL](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1311946)
- [Derrick Johnson](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/bf94f4fc-a760-4403-9065-57178b13dad2) — Qualcomm, Staff Engineer and Developer Advocate
  [Demo: Scaling AI Across Devices: ExecuTorch Portability & Performance from Edge to Mobile, XR, & IoT](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1347503)
- [Digant Desai](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/741f4f81-044b-4c76-a433-c7a689f199a7) — Software Engineer, PyTorch, Meta
  [Building Portable, Composable Local Agents with ExecuTorch](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257786)
- [Driss Guessous](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/d37f545f-612d-402d-900a-3e6f33c3b824) — Meta, machine learning engineer
  [Meet the Developers of PyTorch](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1364540), [Fighting Agents with Agents Bringing Claude to PyTorch CI, triage, and PR review](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1255221), [Observability Tooling for Cudagraph Workloads](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1255975), [FlexGEMM: Flexible PyTorch Epilogues](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1252009)
- [Dunfan Lu](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/02e37759-6364-4986-bac2-f67c808f0033) — Software Engineer - Meta
  [Helion: CuteDSL and TPU Backends for Heterogeneous Hardware, and Why It Suits Agents](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1255964), [Meet the Developers of Helion](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1340815)
- [Dylan Doblar](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/0735b09a-ee9f-4ebf-9366-9ce16339117c) — ML Infrastructure Engineer, NVIDIA
  [Scaling PyTorch Dataloaders 10x: A Case Study](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1243544)
- [Edward Yang](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/842bc307-84ed-4655-891b-55ea92956bb1) — Research Engineer at Meta
  [Keynote: PyTorch Updates](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1298053), [Contributing to PyTorch with AI agents](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1252250), [Meet the Developers of PyTorch](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1364540)
- [Eikan Wang](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/fe44db8a-c695-465f-b248-d58aba846c72) — Intel, AI Framework, Architect, PyTorch
  [PyTorch Generalization: A Journey Toward Write Once, Run Anywhere](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256608), [PyTorch for Agentic AI: Scaling Heterogeneous Systems from CPU to XPU](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1291520)
- [Elfie Guo US](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/6a7a05cb-e141-4922-ab95-52686130aa6e) — Software Engineer, Nvidia
  [Demo Theater: Pretraining at Scale with TorchTitan](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1339821)
- [Elias Ellison](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/87150efb-90fa-4c0e-b18a-377b3d32c283) — Software Engineer at Meta
  [Parametrized Dynamic Shape CUDA Graphs](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257948)
- [Elisa Cascardi](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/310df696-dfb1-4e83-97b0-3d90c7f9d5cb) — Open Source AI @ Meta FAIR
  [How PyTorch Powers Open Foundational Models — From Language to Vision](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256006), [From Neurons to Atoms: How PyTorch Powers Open Science at Scale](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1255173)
- [Emilio Cota](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/6e67bf80-817b-4ce1-b7a3-e6bff2e64227) — Software Engineer, Google
  [Meet the Developers of Helion](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1340815)
- [Eric Karl](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/b95163e6-b090-46d9-955b-83f022a1f932) — CTO, Medikamind Inc
  [From Fragmented AI to End-to-End Radiology Reports: Scalable Clinical LLMs with TorchTitan](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256115)
- [Eric Tang](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/44fb6d9a-45b1-4068-b17e-00a143789294) — Software Engineer
  [SkyRL: Democratizing Scalable RL Training](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1345867)
- [Eric Wang](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/2e18978f-b4a4-40f6-a310-04b469602ef5) — Senior Software Engineer - ML Training Platform @ Pinterest
  [Building a Post-Training Platform for Teams That Don't Own the Training Loop](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257963)
- [Erik Schultheis](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/7c9aa891-33f4-4fc1-9e33-c541889f0697)
  [NVFP4 Training Tips and Tricks](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1347085)
- [Esha Lakhotia](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/d4a9fe64-3005-4ac8-a3a1-c680aea7d078) — Technical Product Manager at AWS
  [From torch.profiler to Hardware Cycles: A Practical Profiling Playbook for AWS Trainium](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257690)
- [Ethan Che](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/9a11a8c2-0892-43f1-a263-b1e27c126cfe) — Software Engineer at PyTorch
  [Smarter Autotuning for Kernels: From Bayesian Optimization to LLM-Guided Search in Helion DSL](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256728), [Meet the Developers of Helion](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1340815)
- [Felipe Mello](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/d712573a-e107-49ca-936d-f596416c1aa3) — Software Engineer at Meta
  [Torchtitan RL: A Unified and Extensible Training Framework for Agentic Tasks](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256870)
- [Felix Baum](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/e5f74b01-6891-4284-b118-29338d78adbd) — Sr. Dir of Product Management
  [Demo: Scaling AI Across Devices: ExecuTorch Portability & Performance from Edge to Mobile, XR, & IoT](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1347503)
- [Ferdinand Mom](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/38602906-ed16-4e66-9c4c-fd6abd9e781c) — Hugging Face
  [Transformers.distributed: Bringing First-Class Distributed Training to 2.4M+ Hugging Face Models](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1253220)
- [Fred Li](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/6ce99363-a930-4c23-aaab-29980bfe7a0a) — General Manager of Computing Open Source Development Team, Huawei
  [Boosting PyTorch’s Heterogeneous Compute Ecosystem: Optimizing LLM Training and Serving on NPU](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1348281)
- [Freya Azad](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/222e35e9-ad53-4793-9f71-107a6ae32571)
  [Meet the Developers of Helion](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1340815)
- [Gaurav Arora](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/38c13539-daef-4458-9e4c-91ecf92f0c8e) — Sr. Staff Software Engineer - ML Training Platform
  [Building a Post-Training Platform for Teams That Don't Own the Training Loop](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257963)
- [Gavin Cohen](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/8ccbbde4-b35e-4835-80c6-7af8fcb74b49) — VP of Product at Clockwork Systems
  [Self-Healing PyTorch Training Through Live GPU Migration](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1322707)
- [George Kurian](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/62a509c4-48a1-4db8-9592-87a50f92f54a) — Distinguished Engineer at Cosmos Lab, NVIDIA
  [From Pixels to Physical Motion: Building World Action Models (WAM) with PyTorch & NVIDIA Cosmos](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257041)
- [George Novack](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/da56a138-7383-4a36-9adc-0873dc2c9cd1)
  [Demo Theater: TPU Megakernels](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1360619)
- [George Novack](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/af0c26a0-b8c3-400c-aa48-d189f64daa9f)
  [Meet the Developers of vLLM](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1340810)
- [Goutam Venhkatramanan](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/b5aedafc-0722-49dc-ad45-415863dc17ba)
  [Meet the Developers of Ray](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1340808)
- [Guray Ozen](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/638645a3-ce23-48be-9f04-b2143852252b) — Principal Compiler Engineer
  [High-Velocity GPU Kernel Authoring with CUTLASS Python](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257904)
- [Haonan Zhang](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/dad2db81-3ce0-4447-98cf-95f82d55140c) — Explorer of AI-native low-precision quantization data formats
  [HiFloat: Democratizing Ultra-Low Precision Training and Inference in PyTorch Ecosystems](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1206584)
- [Huapeng Zhou](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/cb86f2c5-c698-44d5-bb39-d648cd6b0bfc) — BosonAI Member of Technical Staff
  [Scaling Audio AI Infrastructure: Building Voice-Native AI](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1258143)
- [Hubert Banville](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/0ca3e713-b499-4e8c-9c67-dfc8f2659e70) — Research Scientist, FAIR
  [From Neurons to Atoms: How PyTorch Powers Open Science at Scale](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1255173)
- [Huma Abidi](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/682d8530-f2e0-49b5-8b24-a3d71885c191) — Advancing Open AI through Heterogeneous Computing
  [PyTorch for Agentic AI: Scaling Heterogeneous Systems from CPU to XPU](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1291520)
- [Ian Stenbit](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/e0ff3db9-91bf-4e50-aec0-a68d893f878c) — AI Infrastructure Engineer - NVIDIA
  [Scaling PyTorch Dataloaders 10x: A Case Study](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1243544)
- [Ilina Mitra](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/8856f613-27c2-4eed-9908-1852e2d03e21) — Head of Inference Platform at Baseten, former Engineering Manager at Meta
  [Building Highly Efficient Inference System for Recommenders Using PyTorch](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1243435)
- [Ion Stoica](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/982d43d5-7e56-48e9-95b0-220abf8d3e03) — Anyscale / UC Berkeley
  [Keynote: Ray's Evolution to Support AI Workloads](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1298071), [Crossing the Divide: Co-Evolving Kubernetes and Ray for the AI Era](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256376)
- [Itay Alroy](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/1deb8483-494d-47f2-b9f3-eda7a473f160) — Senior Software Architect, NVIDIA
  [Elastic Expert Parallelism in vLLM](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1252137)
- [J Deonarine](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/b70d3e57-481a-4d51-ab11-0bbf97dc46ee) — Distinguished Member of Technical Staff, Dell Technologies
  [From Fragmented AI to End-to-End Radiology Reports: Scalable Clinical LLMs with TorchTitan](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256115)
- [Jacob Szwejbka](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/9d089545-4a4a-4078-a912-49c6190bbba5) — PyTorch Edge, Software Engineer
  [Deploying Robot Policies Across Many Targets with ExecuTorch](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257831)
- [Jago Macleod](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/3ffe5bd3-7b9f-4c6b-b521-9078861c11f5) — Engineering Director, Kubernetes & GKE, Google
  [Crossing the Divide: Co-Evolving Kubernetes and Ray for the AI Era](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256376)
- [Jana van Greunen](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/458fadee-58cf-45e3-9586-a7e8d25d426f)
  [Keynote: PyTorch at Meta - From Internal Workloads to Multi-Silicon Ecosystem](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1301820)
- [JanakiRam Goteti](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/0a88c6cd-b2c8-480d-a769-1bd8afc66a3f) — Senior Staff Software Enginer, Crusoe
  [Practical GPU Programming with Triton for PyTorch Developers](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1252848)
- [Jane Xu](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/bcfc2c5d-dca4-47dc-affa-62de55829f5e) — PyTorch SWE @ Meta
  [Meet the Developers of PyTorch](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1364540), [Clearing the Path Towards an ABI Stable PyTorch C++ Extension Ecosystem](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1252992)
- [Jason Ansel](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/6ac086c7-2c70-4b54-96ad-2e816c314f31) — PyTorch Compilers @ Meta
  [Helion: CuteDSL and TPU Backends for Heterogeneous Hardware, and Why It Suits Agents](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1255964)
- [Jason ANSEL](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/e950888d-313b-4c5a-8bbb-b88151a139fa)
  [Meet the Developers of Helion](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1340815)
- [Jay Dawani](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/22f8a38f-ea63-43a0-958f-bd2108a50948)
  [Demo Theater: Why is Heterogeneous Computing So Hard and Why Does it Have To Be?](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1291533)
- [Jennifer Wei](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/d63a2b8e-4c63-42f1-a230-c3a669e2df73) — Independent AI research engineer focused on training dynamics, optimizer tradeoffs, and reproducible LLM systems.
  [Beyond AdamW: A Practical PyTorch Walkthrough of Muon, Dion, and Orthogonalized Optimizer Variants](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1248446)
- [Jesse Treger](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/72169699-a457-48d4-9871-e0d1932bd9d0)
  [Demo Theater: OpenCLIC: Open Communication for Distributed Inference](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1329720)
- [Jewel K M](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/a10c91bf-5168-4421-ae5e-bbc5805757f4) — Principal Software Engineer in Red Hat PyTorch Engineering team
  [Scaling PyTorch's Compatibility Promise: A Tiered Cross-Repository CI Relay for Out-of-Tree Backends](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1251891)
- [Jiale Huang](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/6ec0261e-906c-4c17-b3b4-30c25faaf659) — Staff Software Engineer at Crusoe
  [Demo: Fitting Large MoE Fine-Tuning onto Few Nodes: Expert & Context Parallelism in DeepSpeed](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1317873)
- [Jiani Wang](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/57581580-9cf2-451a-b214-180cdd249f02) — Software Engineer, PyTorch
  [Torchtitan RL: A Unified and Extensible Training Framework for Agentic Tasks](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256870)
- [Jongsok Choi](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/dabfb46d-41d5-42e2-8459-b0b96bd5f983) — Engineering Manager, Meta
  [Smarter Autotuning for Kernels: From Bayesian Optimization to LLM-Guided Search in Helion DSL](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256728)
- [Jongsok CHOI](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/c3d5400e-3892-47b7-805b-37b45429e25e)
  [Meet the Developers of Helion](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1340815)
- [Joseph Groenenboom](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/43e8b706-5c26-4878-95fe-6ea056758e68) — Frameworks engineer invested in open source access to AI
  [Making Enterprise Agentic Inference Production-Ready with PyTorch and vLLM](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1301809), [Ecosystem Working Group: Unlocking Community Impact Through the PyTorch Landscape](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1253136)
- [Joshua Lee](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/c28e3210-1943-44ec-a442-e6503060d739) — Ray Core SWE
  [Demo Theater: Evolving Ray Core for Post-Training at Scale](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1345871)
- [Juan Muneton Gallego](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/86ea42d0-1d67-4291-bcb5-8aaa910d561a) — Intel Corporation, AI Software Solution Engineer
  [From Weeks to Overnight: Autonomous Day-0 Kernel Bring-Up with Agent Pipelines](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256713)
- [Kaiming Cheng](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/b33fc27e-2735-4035-a83f-39b7790e9e44) — Meta SuperIntelligence Labs - PyTorch Team| Research Scientist
  [KernelAgent: Hardware-Guided GPU Kernel Optimization via Multi-Agent Orchestration](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1251908)
- [Kapil Sharma](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/77fd5103-92a9-4c3e-a919-726a80e8f4bd) — Kapil Sharma, Meta, PyTorch Distributed
  [Future of Distributed Communication in PyTorch: New APIs for Fault Tolerance, RDMA and Extensibility](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256220), [Beyond Collectives: Building Fused Compute-Communication Kernels with PyTorch Symmetric Memory](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257030)
- [Karthick Panner Selvam](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/4ebe68c7-a121-411a-b28a-6354981c0179) — SWE at Meta FAIR
  [Meet the Developers of Helion](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1340815)
- [Kavya Sri Chennoju](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/e1f1b1e0-9642-47b6-8991-50c09b3549c9) — Staff AI Engineer at Arm
  [Hardware-Aware AI: Building Agentic Systems from Cloud to Edge with PyTorch, ExecuTorch](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1291518), [Demo Theater: From Prompt to Physical Action: A Live Hardware-Aware AI Demo with PyTorch, ExecuTorch](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1291522)
- [Ke Chen](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/5c1d292a-6d79-4403-95cb-7f137f329924) — Staff Software Engineer at Uber
  [PyTorch-Native Feature Transformation and Training Framework for Uber Eats Recommendation](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1241540)
- [Ke Wen](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/f9b10346-995d-4c20-a3a4-fdabef35fb13) — NVIDIA, Principal Software Architect, PyTorch Maintainer
  [Beyond Collectives: Building Fused Compute-Communication Kernels with PyTorch Symmetric Memory](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257030)
- [Kevin Li](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/ee6fa020-ab7c-437b-ad0d-bd80a6d62c4b) — Member of Technical Staff, Verda
  [Demo Theater: WideEP NVL72 with Rust Frontend](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1361627)
- [Kevin Mi](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/a52f56e3-4245-4f28-9258-e8bbe32bd63f) — Member of Technical Staff, RadixArk
  [SGLang-Diffusion: High Performance Serving Framework for Offline and Real-time Diffusion Models](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1258314)
- [Khush Patel](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/0ce0647c-fe89-4e73-aad5-ed0a40f699bd) — Computer Science Graduate & Open Source Contributor, SimplexLab
  [TorchJD: Training PyTorch Models with Multiple Losses](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256805)
- [Khushboo Rathi](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/c764ac31-e3cf-44bb-a823-a9b1499f3711) — AI Engineering Technologist
  [From Fragmented AI to End-to-End Radiology Reports: Scalable Clinical LLMs with TorchTitan](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256115)
- [Kuntai Du](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/8ef1af70-b6b3-4036-a4d4-22e031905de9) — Co-founder and Chief Scientist of Tensormesh. vLLM and LMCache committer.
  [LMCache: a cluster-wide open-source solution for LLM prompt caching](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256217)
- [Laith Sakka](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/d6c65b7a-04c4-4094-96c9-9cdb27c58f52) — Meta
  [From Backed to Unbacked: Sound, Predictable, and Controllable Dynamic Shapes in PyTorch](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1260092)
- [Laura Gustafson](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/2ae55485-cd4a-4627-8335-3a2cf35c7f55) — Research Scientist at FAIR, Meta
  [How PyTorch Powers Open Foundational Models — From Language to Vision](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256006)
- [Laura Wang](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/2d6077f4-3da7-43b6-bd0d-eea01cef0486) — MSL-Pytorch | Software Engineer
  [KernelAgent: Hardware-Guided GPU Kernel Optimization via Multi-Agent Orchestration](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1251908)
- [Liana Koleva](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/61bff2d5-d360-4e64-827a-1f180c2eba4e) — Research @ Lightning AI
  [What Would a PyTorch Expert Say? Turning Tribal AI Knowledge into a Model](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1333197)
- [Lin She](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/c655aff8-f949-4e62-8e6a-3250a18c7928) — Software Engineer at Clockwork.io
  [Demo Theater: Never Hit Zero: A Live Demo of Surviving Link and Node Failures in PyTorch Training](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1322762)
- [Lindsey Allen](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/361c4162-0c14-4d64-a555-fa88a0a74d2e) — Boson AI Turning Voice interaction into business advantage
  [Scaling Audio AI Infrastructure: Building Voice-Native AI](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1258143)
- [Liz Li](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/314168a6-1e01-44ac-a0f8-587729655676) — AI architect in AMD
  [Extending TorchInductor with FlyDSL: A New MLIR-Native Backend for High-Performance GEMMs](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256835), [MXFP8/4 pretraining at scale with Torchtitan on AMD Instinct GPUs](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256822)
- [Lu Fang](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/0908c6d5-aa13-45f0-a30f-c76892078235) — Research Scientist at Meta
  [Building Highly Efficient Inference System for Recommenders Using PyTorch](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1243435)
- [Luc Georges](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/5c6e46ee-f9ca-4027-ba72-019619329190) — Hugging Face, ML & Software Engineer
  [Meet the Developers of Safetensor](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1340758)
- [Lucas Wilkinson](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/5064daaf-674a-44ac-b20d-084a6512880d) — vLLM Core Maintainer | Senior Principal Engineer RedHat
  [A Developer’s Guide to Attention in vLLM](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1254038)
- [Maen Suleiman](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/06cdc1ea-d5dd-416c-89bb-f4f1c79ead01) — Principal Product Manager, AWS Neuron
  [Training on AWS Trainium via TorchNeuron: TorchTitan and Hugging Face Transformers v5](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256874), [Demo: PyTorch Ecosystem Running Natively on Trainium: Reinforcement Learning - TorchTitan & Monarch](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1291526)
- [Mahdi Ghodsi](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/f4d2f21c-6e31-4aee-84ef-74ed75721724)
  [Building Cost-Efficient Mixture-of-Models on AMD GPUs with vLLM Semantic Router](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1333195)
- [Marie Antons](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/810414bf-0260-4129-9557-133e84eff1a3) — Enterprise Sales Engineer, Datadog
  [Demo Theater: Where’s My Training Job?](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1314428)
- [Mark Collier](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/79669461-8189-4395-8b93-1aeb5f651f95)
  [Keynote: Accelerating Open Intelligence Together](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1298042), [Keynote: Building What’s Next in Open Source AI](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1303522)
- [Mark Moyou](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/95b932ec-0f30-44c9-b365-e01d76a14138) — Sr. Solutions Architect - NVIDIA
  [Understanding Modern Vision Language Models](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1260323)
- [Mark Saroufim](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/a8fa0d41-dfbe-4e52-895c-44139ed8cba8)
  [Keynote: Linear Algebra for the Age of Research](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1298055)
- [Markell Rawls](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/76e71d7d-998c-48a7-8ef9-f5680795ba5c) — Technical Marketing Engineer, Red Hat
  [Demo Theater: Quantization Showdown: PyTorch Inference Optimization](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1291525)
- [Maroon Ayoub](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/536bb926-c468-4f04-9f0f-c79565256dc2) — Senior Principal Machine Learning Engineer, Red Hat
  [Contributing to Inference OSS That Won't Stand Still: A BoF on vLLM, llm-d, and the Moving Target](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257610)
- [Masahiro Tanaka](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/c9095c0b-90d0-42bd-8756-b466acb78617) — Anyscale
  [Meet the Developers of DeepSpeed](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1364544), [DeepSpeed Is Not Just ZeRO: Tensor, Sequence, and Expert Parallelism with ZeRO-Level Usability](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257990)
- [Matt Cossins](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/3879c6e4-ea8e-4d15-a7df-2033c2b9292d) — Developer Ecosystem Manager, AI & Developer Platforms, Arm
  [“What Do I Do Now?” Bridging the ExecuTorch Adoption Gap Through Developer Learning](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1255970)
- [Matthew Bonanni](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/485337ce-dc62-4af4-8e31-3e5ad923da5d) — vLLM Maintainer | MLE at Red Hat
  [A Developer’s Guide to Attention in vLLM](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1254038)
- [Mayank Debnath](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/9dbc40a0-a7da-42c0-b01f-37c3c24107bd)
  [Demo Theater: Your AI Model Needs More Than an API](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1362826)
- [Mazin Gilbert](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/d3a94032-182e-4190-9516-4f1c509b4bf2)
  [Keynote: From AI to Agentic AI and Back](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1298047)
- [Mengjin Yan](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/d190c3d7-5cdd-4050-911a-97a690f83e32) — Engineering Lead, Ray Core, Anyscale
  [Demo Theater: Evolving Ray Core for Post-Training at Scale](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1345871)
- [Mengyin Yan](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/b6277307-1122-430c-8037-876ce846dfd5)
  [Meet the Developers of Ray](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1340808)
- [Mergen Nachin](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/dbe26457-34cb-45fd-9f0d-1ec15900aedf) — Meta, Software Engineer
  [Building Portable, Composable Local Agents with ExecuTorch](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257786)
- [Michael Benayoun](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/9393eea6-9167-4702-9f0c-e37e954be0f5) — Machine Learning Engineer @ Hugging Face
  [Training on AWS Trainium via TorchNeuron: TorchTitan and Hugging Face Transformers v5](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256874)
- [Michael Goldfarb US](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/44799cba-6235-4c38-95dd-55784e352185) — NVIDIA
  [High-Velocity GPU Kernel Authoring with CUTLASS Python](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257904)
- [Minjia Zhang](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/b373a478-4988-4f87-9916-8562a9e63666)
  [Meet the Developers of DeepSpeed](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1364544)
- [Minwook Ahn](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/2a775773-5dda-4800-b312-9b12c0839d36) — Rebellions, Senior Developer
  [Demo Theater: Beyond torch.compile: Reducing Data Movement with Device-Persistent Tensors in PyTorch](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1291535)
- [Mu Li](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/87142268-038a-4a91-bd64-cd8cb53d2118) — BosonAI CTO
  [Scaling Audio AI Infrastructure: Building Voice-Native AI](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1258143)
- [Mudhakar Srivatsa](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/ec29e2c3-cb8a-49f1-aec9-0a5058771a4d) — IBM
  [Demo Theater: IBM Spyre Accelerator in The PyTorch Open Ecosystem](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1351816)
- [N Maajid Khan](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/00138e96-90de-43d5-9989-f60be40a383c) — MONAKA SW R&D, Lead Software Engineer, Fujitsu Research of India Pvt. Ltd.
  [Efficient MoE LLM Inference on Arm with vLLM and OpenVINO](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257004)
- [Natalia Gimelshein](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/cf9b7b36-8fcb-4506-9215-566c6bc442d0) — Meta
  [Meet the Developers of PyTorch](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1364540), [Observability Tooling for Cudagraph Workloads](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1255975), [Beyond Collectives: Building Fused Compute-Communication Kernels with PyTorch Symmetric Memory](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257030)
- [Neiwen Ling](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/ffaa2c45-f744-4014-8d88-077c41d42ef7) — ByteDance, Research Scientist
  [When Rollout and Training Disagree: Lessons from Building a Mismatch-Free RL Engine](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257767)
- [Nick Hill](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/c5c6a2fd-9c77-4fbc-a221-be26eec2b36e)
  [vLLM Roadmap](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1355211), [Meet the Developers of vLLM](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1340810)
- [Nicolas Hug](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/ac372e6f-acc1-432c-ae3d-12cc311d0741) — Meta - Research Engineer
  [From Bytes to Tensors: Efficient Media Processing with PyTorch in 2026](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1254966)
- [Nicolò Lucchesi](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/c44bac6e-1d65-4d15-9729-5dfc693a1a89) — Research Engineer, Mistral AI, vLLM Maintainer
  [State-of-the-Art KV Transfer for Disaggregated LLM Serving in vLLM](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1255023)
- [Nili Guy](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/7cdc8e3e-f9ec-49c5-80a3-81f92b0a10ce) — IBM Research Manager
  [Contributing to Inference OSS That Won't Stand Still: A BoF on vLLM, llm-d, and the Moving Target](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257610)
- [Oguz Ulgen](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/643442a4-624e-4b8b-8a1f-9837dfb69605) — Software Engineer, Meta
  [Helion: CuteDSL and TPU Backends for Heterogeneous Hardware, and Why It Suits Agents](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1255964), [Meet the Developers of Helion](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1340815)
- [Olivier Tardieu](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/92419682-e50b-409b-b1dd-f2557bb3d3bf) — Principal Research Scientist, Manager at IBM Research
  [Beyond Size and Stride: Unleash Performance with Device-Aware Tensor Layouts](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256241)
- [Or Ozeri](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/518cd396-541e-4f05-a8fd-4c2491d1cb8f) — IBM, Cloud Researcher
  [Native Tiered KV Cache Offloading in vLLM: From Storage Offloading to Disaggregated Serving](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1254938)
- [Oscar Savolainen](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/6f310554-989c-407d-9022-5d2d310e227e) — Dedicated Inference at Nscale
  [Lessons at Scale: Building a PyTorch Inference Stack with Open-Source Components](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1347541)
- [Panagiotis Kourdis](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/d7cf2f41-f774-4701-b76d-f8329bbdcc73) — AI Frameworks Engineer, Intel
  [XCCL: Scaling PyTorch Collectives to Exascale on Intel GPUs with TorchComms](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1258012), [Scaling PyTorch on Intel XPU: Field Notes from AI for Science on Aurora](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257912)
- [Paul Chang](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/5a08b8d9-e579-4c08-bb1b-c3ba0fcf3a61)
  [NVFP4 Training Tips and Tricks](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1347085)
- [Paulo Aragao](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/053d32d2-63dd-4815-a8dd-8136dcccff1c) — Principal, Applied AI Architect, AWS
  [Five Bottlenecks, Five Fixes: Systematic Performance Diagnosis for PyTorch Distributed Training](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1254803)
- [Pavneet Ahluwalia](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/2e10b79f-062a-4a53-8634-92610cbb703a) — Product @ Microsoft Azure
  [Self-Healing PyTorch Training Through Live GPU Migration](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1322707)
- [Peng Zhang](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/84e468f7-18b1-4fc5-8fc9-4c545416209a) — Engineering Manager
  [PyTorch-Native Feature Transformation and Training Framework for Uber Eats Recommendation](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1241540)
- [Peter Prettenhofer](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/26ab0f92-2c44-4156-b4c2-b8de1e3e6313) — Robotics Machine Learning Engineer at Rerun
  [Multimodal Dataloaders for Physical AI](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257585)
- [Peyman Razaghi](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/34ee5947-3376-42c2-a774-2594a034b5bd) — Peyman Razaghi, PhD
  [Primus Tuning: Hybrid Projection and Agentic Search for Distributed Training](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256459)
- [Philip Kiely](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/7cb31c6c-fcf2-4d0a-b4bb-6461d7776369) — Head of Developer Relations at Baseten
  [What's New in Inference in October](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1317774), [Book Signing with Philip Kiely](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1354949)
- [Pian Pawakapan](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/f0ffa860-0c87-45f7-949e-92587cea5003) — Software Engineer, PyTorch Compiler/Distributed
  [From Full DTensor to spmd\_types: Making Parallelism Explicit in TorchTitan](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257076)
- [Pinak Panigrahi](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/988ea39f-21d3-44e1-afff-3637a2c8069e) — Annapurna ML, Amazon
  [From torch.profiler to Hardware Cycles: A Practical Profiling Playbook for AWS Trainium](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257690)
- [Prachi Gupta](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/afc23c94-0247-4201-a2cc-b78853bd9ac3) — Engineering Manager, PyTorch Dev Team @ AMD
  [rocSHMEM Symmetric Memory in PyTorch for AMD GPUs](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256215)
- [Prasanth Chatarasi](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/91842132-764e-4860-a334-bde0aa299888) — Senior Research Scientist (IBM Research)
  [Unlocking PyTorch for Dataflow Accelerators: An Open-Source Kernel Tile IR and Dataflow Scheduler](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257807)
- [Priya Nagpurkar](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/70be2ac1-0d90-4538-9095-7cf971bd2a53)
  [Keynote: From Silicon to Models at Scale](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1298073)
- [Purva Chiniya](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/f3b44840-a999-4c95-acef-bee81d34eeef) — Applied Scientist at Amazon Ads | Secure Agentic AI & Multi-Agent Systems
  [Compute, Latency, and Safety: Architecting Stateful Guardrails for Multi-Agent Workflows](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1251388)
- [Qi Zhou](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/f651ba85-1c3d-40e7-8458-25ed8f51e5f7) — Software Engineer, Google
  [PyTorch-Native LLM Serving on TPU: SGLang and vLLM](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256716)
- [Qiacheng Li](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/9bc03284-808b-49f4-9c16-8291d952e529) — Consumer AI Technical Marketing Engineer, Intel
  [Demo Theater: AI Playground Home Agent: Remote PyTorch AI Workflows from Your Phone](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1291528)
- [Rachad Alao](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/c35d5595-a046-45a6-af81-c57ce473726a)
  [Keynote: Open Science and AI Foundation Models: Allies in the Pursuit of Knowledge](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1298057)
- [Rain Jiang](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/ebb723f3-d0b6-4043-89aa-aff972a782ec)
  [Demo Theater: WideEP NVL72 with Rust Frontend](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1361627)
- [Ravi Teja Prabhala Venkata](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/4dcedc69-567a-4342-bca7-9e9da8fd8f0f) — Senior Engineering Manager
  [Closing the Confidence Gap: Making AI Output Trustworthy in Enterprise Systems](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256339)
- [Ray Gao](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/a0214499-6b3f-434a-a3e7-61574ea1ac33) — Research Engineer at FAIR Chemistry
  [From Neurons to Atoms: How PyTorch Powers Open Science at Scale](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1255173)
- [Ricardo Noriega](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/4dac1a96-71c9-4537-b5f3-e46ba65fa11f) — Principal Software Engineer, Edge Computing Team, Office of the CTO - Red Hat
  [Prefix Caching for Autoregressive Stages in Multi-Stage Pipelines](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1255511)
- [Richard Liaw](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/76112655-3827-4f31-afe6-b33cbffc6397)
  [Meet the Developers of Ray](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1340808)
- [Richard Zou](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/8080a67e-84a3-42a9-88ab-7827903b974b) — Software Engineer @ PyTorch
  [One Model Definition, Many Accelerators: Scaling vLLM Across Hardware Without Forks](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257448), [Lightweight FX Tracing in PyTorch](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256183)
- [Rob Mulla](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/9286174a-7239-4df9-b4ce-4caefe4cc486) — ML Frameworks @ Google
  [Demo Theater: Unifying Open-Source LLM Serving on Google Cloud TPUs with TorchTPU](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1304122)
- [Robert Hundt](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/bdaf8c0c-8999-4f8c-a5e9-b0ac90c3b5b7)
  [Keynote: Trainium’s Journey to Native PyTorch](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1298065)
- [Robert Nishihara](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/7ba0e0db-b8be-4a8d-a88e-b410b88ec162) — Co-founder, Anyscale
  [Meet the Developers of Ray](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1340808)
- [Ruijie Zheng US](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/8a80ceec-7890-42a8-91e0-ac223ed5a04a) — Research Scientist at NVIDIA GEAR Lab
  [From Pixels to Physical Motion: Building World Action Models (WAM) with PyTorch & NVIDIA Cosmos](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257041)
- [Ryan Spring US](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/cfb7bdef-133f-4396-bd99-f03fa7214abf) — Software Engineer
  [Efficient Pretraining of LLMs in NVFP4](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256332)
- [Sahdev Zala](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/2592a0c3-4592-48cf-9a30-c5c59cf01d73) — Senior Technical Staff Member, Open Source Developer
  [Inside the PyTorch Ambassador Model: How Ambassadors Grow PyTorch Adoption & How You Can Become One](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1306148)
- [Sai Charan Teja Gopaluni](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/98098927-a120-47ec-b30f-969ce7b14956) — Amazon Web Services, Sr. Specialist SA, GenAI & Agentic Platform
  [Guardrails at the Infrastructure Layer: Sandboxing and governing Agentic AI in production](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1244699)
- [Sam Foreman](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/a34d2a4c-1d88-40fe-9988-da7bbb964faf) — Computational Scientist
  [Meet the Developers of DeepSpeed](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1364544), [Scaling PyTorch on Intel XPU: Field Notes from AI for Science on Aurora](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257912)
- [Sandeep Pokkunuri](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/2cd5c678-4400-4bed-af71-6546cf712470)
  [TPU agent suite: Accelerating PyTorch Model Onboarding and Inference on Cloud TPUs](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1301802)
- [Sanket Jayant Purandare](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/08dcdb2a-60ca-496f-9e0e-7fb3efd6868c) — Research Engineer, PyTorch Distributed and Compilers
  [EP-Overlap: Hiding Expert-Parallel Communication in MoE Training](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257875), [Dist-MoE: Fused Comm-Compute Kernels & Activation Packing for Speed-of-Light MoEs](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257887), [TorchTitan GraphTrainer: LLM Distributed Training with PyTorch's Compiler Toolkit](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257898)
- [Sara Hooker](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/36d529c4-c92c-44d2-b956-afb640884778)
  [Keynote: Beyond Brute Force: The Era of Adaptive Intelligence.](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1298060)
- [Sarah Mohamed](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/4a6edfc0-8d09-4b77-bf22-cd50e701f207) — Principal Software Engineer, MathWorks
  [Deploying PyTorch Models Where Failure Isn't an Option](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1251868)
- [Scott Schneider](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/5cb271c8-e727-437f-9c69-fe7688079fa2) — Meta - Software Engineer
  [From Bytes to Tensors: Efficient Media Processing with PyTorch in 2026](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1254966)
- [Sean McGovern](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/670fab87-ba88-47b7-a3f3-30ed306281b3) — PyTorch Engineering @ Red Hat
  [Clearing the Path Towards an ABI Stable PyTorch C++ Extension Ecosystem](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1252992)
- [Sergio Paniego Blanco](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/61d7a90f-9744-4ad8-bb0d-34e5961881e6) — Hugging Face - Machine Learning Engineer
  [Train the Agent, Not Just the Model](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1255556)
- [Shangdi Yu](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/65629c5a-53d3-4b69-933d-789147477236) — Meta
  [Meet the Developers of Helion](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1340815)
- [Shashank Verma](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/d3ca9877-ec10-4922-90d5-98c1500fc84c) — Product Research Manager - Deep Learning at NVIDIA
  [The Nemotron Ecosystem: Building the Third Generation — Nano, Super, and Ultra](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256810)
- [Shekhar Pandey](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/6364d80d-2773-4b47-8fad-a4432f5102e3) — Member Of Technical Staff at AMD
  [MXFP8/4 pretraining at scale with Torchtitan on AMD Instinct GPUs](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256822)
- [Sheng Huang](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/de23d555-b39f-418a-bc40-59d20ae187d6) — Software Engineer, ML Training at Pinterest
  [Scaling Foundation Models: From Broken to Near-Linear Multi-Node Training](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256382)
- [Sherlock Huang](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/dfc3bea0-6eb4-4364-9e04-843b90176f4a) — Meta, PyTorch Compilers, Software Engineer
  [TorchTitan GraphTrainer: LLM Distributed Training with PyTorch's Compiler Toolkit](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257898)
- [Shikai Li](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/e8b610f1-c03b-4018-b657-6a18f584e4bd) — Software Engineer in Meta SuperIntelligence Labs
  [Dist-MoE: Fused Comm-Compute Kernels & Activation Packing for Speed-of-Light MoEs](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257887)
- [Shuhua Yu](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/b0a3f727-8fa2-42cb-9e6a-ec3d1e3ba56a) — Research Scientist, PyTorch
  [Agentic RL Training in Pytorch](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256335)
- [Shunyao Li](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/e767d689-b701-4fce-85ee-ff7ed6c5c48c) — Senior Software Engineer - ML Training Platform @ Pinterest
  [Building a Post-Training Platform for Teams That Don't Own the Training Loop](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257963)
- [Simon Layton](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/82624e29-70d2-4efe-8678-145a05f91ca4) — PyTorch Core Engineer
  [Native DSL Operators in PyTorch Core](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1255976)
- [Simon Mo](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/a91796a5-1add-40b0-bb41-86ac45556efa)
  [Keynote: vLLM Update: Scaling Open Frontier Inference Infrastructure](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1298061)
- [Sol Kim](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/9f973733-5340-451a-89d1-bba09bc0a8b0) — FuriosaAI, Software Engineer
  [Demo Theater: Composable Tensor Programming Across Compiler Stages: PyTorch, TCL, and VISA](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1342322)
- [Songhao Jia](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/1ce6a6b9-d707-4aed-9d88-c1e2fc4aa1e8) — Software Engineer at Meta PyTorch
  [How PyTorch Powers Open Foundational Models — From Language to Vision](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256006)
- [Sonny Mupfuni](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/4a8d56e0-28d0-4694-b605-b9c6343375b2) — AI Engineer
  [Inside the PyTorch Ambassador Model: How Ambassadors Grow PyTorch Adoption & How You Can Become One](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1306148)
- [Sreeram Potluri](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/1d07c761-325e-4979-908d-ccda20fc7b61) — NVIDIA, Distinguished Software Architect
  [Introducing NCCL Extensions: Communication Patterns for Modern AI](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257023)
- [Stefan Lindall](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/eff1d986-10ab-4a34-98e3-6bf26264ff09) — Making AI Fast
  [JIT Kernel Compilation: How Modular writes fast kernels for any hardware](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257411)
- [Stephanie Wang](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/cb052273-cd84-4389-8081-27a37e168b06) — Creator of Ray, Assistant Prof at UW CSE
  [Unlocking the Full Potential of TorchDynamo: Accelerating, Comparing, and Debugging ML Systems](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257418)
- [Stephen Watt](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/8c5235a9-dbf7-4b4c-a3dc-f6dcbada30be)
  [Keynote: Open Source Inference at Production Scale](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1302854)
- [Steven Pousty](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/cacf8376-f428-4c3d-81fb-180f2f665065) — Principal Community Architect
  [Gain Geospatial Understanding with the PyTorch Ecosystem](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1258269)
- [Steven Troxler](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/f30649b5-0764-4f3b-bcf3-637447b6eca8) — Python Language Tools at Meta
  [Static Tensor Shape Checking for PyTorch with Pyrefly](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256337)
- [Subin George](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/339522e6-6cbe-4d19-8a86-e4970ded7bc9) — Senior Software Engineer, PyTorch Engineering, Red Hat
  [Scaling PyTorch's Compatibility Promise: A Tiered Cross-Repository CI Relay for Out-of-Tree Backends](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1251891)
- [Sudhanshu Shrivastava](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/9f5b826c-bf89-4164-b259-e609a3574e24) — Graduate Student, Electrical and Computer Engineering Department, UC Davis
  [Privacy-Preserving AI Across Institutions: Federated Learning Lessons for Medical Imaging](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1199496)
- [Suman Debnath](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/13e70062-52c2-4c2d-8b1e-8ccbcec6917c) — Director, Developer Relations and Product
  [Practical GPU Programming with Triton for PyTorch Developers](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1252848), [Demo: Fitting Large MoE Fine-Tuning onto Few Nodes: Expert & Context Parallelism in DeepSpeed](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1317873)
- [Sumanth Hegde](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/99913324-ed84-4925-a7e3-e11fe4d5ed73) — Software Engineer, Anyscale
  [SkyRL: Democratizing Scalable RL Training](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1345867)
- [Sunita Nadampalli](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/31403442-97e0-4ecf-a4b0-a46b13af65f6) — Principal Engineer, AI/ML expert at Amazon
  [State-of-the-Art KV Transfer for Disaggregated LLM Serving in vLLM](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1255023)
- [Susie Xia](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/d85df0f6-aeec-4801-a524-c677826aa006) — Principal Engineer, NVIDIA
  [From Pixels to Physical Motion: Building World Action Models (WAM) with PyTorch & NVIDIA Cosmos](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257041)
- [Suvaditya Mukherjee](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/357b510e-528a-4458-9869-f0e363e09833) — ML @ Hugging Face | MS AI @ USC | ML Google Developer Expert
  [What You Cannot Profile, You Cannot Optimize: Learning to Read PyTorch Traces](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1255448)
- [Swagath Venkataramani](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/3b61972c-caae-4400-bb50-1b7e91660e3e) — Principal Research Scientist, IBM Research
  [Unlocking PyTorch for Dataflow Accelerators: An Open-Source Kernel Tile IR and Dataflow Scheduler](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257807)
- [Syed Ahmed](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/a7e5e154-5275-4a0e-a7e9-58e202dfc364) — Senior Software Engineer, PyTorch Core @NVIDIA
  [EP-Overlap: Hiding Expert-Parallel Communication in MoE Training](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257875)
- [Takeshi Yoshimura](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/092d4cd8-736a-48a9-bded-6a754e15239b) — IBM Research - Tokyo
  [Faster LLM Serving Startup with fastsafetensors](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1255212)
- [Tanima Dey](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/ea8dcaa8-51ac-4b72-94ae-b9f3ca2ec499) — AI Frameworks Engineer, Intel
  [XCCL: Scaling PyTorch Collectives to Exascale on Intel GPUs with TorchComms](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1258012), [Scaling PyTorch on Intel XPU: Field Notes from AI for Science on Aurora](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257912)
- [Tao Huang](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/14cfd922-f4ff-447a-80ed-35785e0b4720) — Senior Software Engineer, Systems Infrastructure at LinkedIn
  [Ray All the Way Down: A Heterogeneous, Elastic PyTorch Training Stack at LinkedIn](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257598)
- [Teja Pulagam](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/7e4eace3-e01f-47c0-a725-15e39807c7f9) — Research Engineer, Lightning AI
  [What Would a PyTorch Expert Say? Turning Tribal AI Knowledge into a Model](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1333197)
- [Theotime Combes](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/1f14068b-e6f4-49f9-89bd-b90178a78447)
  [Meet the Developers of Helion](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1340815)
- [Thomas Cottenier](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/687c9568-bc68-466a-adcb-4e1bf2239ccb) — Arm, Senior Principal Engineer
  [From PyTorch to the Edge: Agentic Synthesis of Inference Runtimes for Heterogeneous Hardware](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257310)
- [Thomas Parnell](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/7e2d6eef-4686-4de7-b671-fe051ea82242) — Principal Research Scientist @ IBM
  [One Model Definition, Many Accelerators: Scaling vLLM Across Hardware Without Forks](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257448)
- [Thomas Wang](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/f51d4db4-100b-4431-ace7-2621f62c483c) — Software Engineer, Meta
  [Monarch - Single-Machine DevX at Distributed Scale](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1255147)
- [Tianle Zhong](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/7e0f5a2b-feeb-446c-88d7-ebf12e6fbfd2) — PhD student at University of Virginia
  [When Rollout and Training Disagree: Lessons from Building a Mismatch-Free RL Engine](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257767)
- [Tommy Li](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/9fa123e5-0f25-40ea-ac37-9bd14772ccfa) — Staff Software Developer
  [Ray All the Way Down: A Heterogeneous, Elastic PyTorch Training Stack at LinkedIn](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257598)
- [Trenton Starkey](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/eddf7b7e-d568-42cd-bdd3-34f801ae7cc3) — NVIDIA Product Manger
  [Fast, Small, Accurate: Optimizing Models for Inference using FP4 Quantization](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1253987)
- [Trinadh Kotturu](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/5d55fc9a-f908-45ff-83dc-ef43a9b12207) — Senior Product Manager, Google
  [Keeping GPUs Busy: High-Speed Storage for PyTorch via fsspec](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1228462)
- [Tristan Rice](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/dde599d6-57ee-4fbc-9a07-ae71ff0c3c61) — Meta - PyTorch
  [Meet the Developers of PyTorch](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1364540), [Future of Distributed Communication in PyTorch: New APIs for Fault Tolerance, RDMA and Extensibility](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256220)
- [Tristan Webb](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/c13802d5-467b-4d59-af59-f0b7a1d557ac) — ML Engineer and K8 admin at d-Matrix
  [Demo Theater: dmx-compressor: Accelerating the Development of Kernels for Custom ASIC Hardware](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1309107)
- [Tunji Ruwase](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/20c4d7bd-1fa1-4d06-8fb1-21f9e64be97e)
  [Meet the Developers of DeepSpeed](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1364544)
- [Tyler Michael Smith](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/0366c21c-0dbf-48d9-b284-88c152301389) — Chief Architect - Inference Engineering @ Red Hat
  [Making Enterprise Agentic Inference Production-Ready with PyTorch and vLLM](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1301809)
- [Ujval Kapasi](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/2a2b6d7e-0d7e-4750-9186-31699c68a39d)
  [Keynote: PyTorch Performance for Agents on Vera Rubin](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1298063)
- [Valérian Rey](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/f41042d8-203a-4e57-bb98-72864800de37) — Simplex Lab, Research Engineer
  [TorchJD: Training PyTorch Models with Multiple Losses](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256805)
- [Vijay Janapa Reddi](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/4612b91c-5148-42c1-a635-d51564aa0dc6) — Professor, Harvard University | Author, Machine Learning Systems
  [From Scratch to PyTorch: Demystifying ML Frameworks by Building Your Own](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256633)
- [Viji Srinivasan](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/3cb1cc6d-c5a2-46fe-be42-df69b25b706a) — IBM (Research)
  [Unlocking PyTorch for Dataflow Accelerators: An Open-Source Kernel Tile IR and Dataflow Scheduler](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257807)
- [Wei Feng](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/4acfc150-3647-41ba-b6a3-e3ae788726d2) — Software Engineer, PyTorch Distributed at Meta
  [FlexShard: Pluggable Sharding Placements for Matrix Optimizers and Block-wise Quantization](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256015)
- [Whitney Tsang](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/bd0df34b-9097-45d2-b25a-9dc42d0593d3) — Principal Engineer @ Intel
  [Making vLLM Faster on Intel GPUs with Triton Kernels](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1255903)
- [Will Constable](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/01efcdf1-0f41-462f-8cf0-a689b078c6c8) — PyTorch Distributed at Meta
  [Meet the Developers of PyTorch](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1364540), [Integrating TorchTPU into TorchTitan and Optimizing Models for TPUs](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1246272)
- [William Wen](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/f520a9df-e576-453a-bcc9-efd16733a557) — Software Engineer, Meta
  [Nested Graph Breaks: Reducing the Cost of Graph Breaks in torch.compile](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256035)
- [Xandra Zhu](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/502e3612-f1c1-473d-8821-1acfc5c767c8) — Uber, Machine Learning Engineer
  [PyTorch-Native Feature Transformation and Training Framework for Uber Eats Recommendation](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1241540)
- [Xiaohu Guo](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/5af35e9b-8b49-4235-b03d-ddbf3a0b7c84) — SMTS, AMD
  [PerfModel: A Validation-Driven Performance Model for Triton Kernels](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257198)
- [Xiyuan Wang](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/6b5d35dd-d75f-4e76-908a-bc21bd7d772e) — vLLM Ascend Maintainer
  [vLLM KV Cache Management for Model-Specific Requirements](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1253418), [Advancing torch.compile for Verifiable Precision & Dynamic Shapes](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1253426)
- [Yahav Biran](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/b367b1ef-cc31-42ea-8e25-9509ee815777) — Amazon Web Services, Principal Architect
  [TorchNeuron: Native PyTorch RL on AWS Trainium - From Research to Production Without Compromise](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1248603)
- [Yarong Mu](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/3945da87-19f4-41ac-83ec-22deb3131983)
  [Meet the Developers of Helion](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1340815)
- [Yi Pan](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/3163ec85-964d-44ab-b3a7-7e4d60898213) — PhD Student at UC Berkeley Sky Computing Lab
  [Unlocking the Full Potential of TorchDynamo: Accelerating, Comparing, and Debugging ML Systems](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1257418)
- [Yichuan Wang](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/d9b0dcc5-eb21-4e7a-be3a-01b967aeace8) — Meta | PhD Student @ UC Berkeley
  [Agentic RL Training in Pytorch](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256335)
- [Yidi Wu](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/a233ea6a-7e27-4c60-b7fe-417c8ff97a27) — Meta, PyTorch Compilers, Software Engineer
  [Lightweight FX Tracing in PyTorch](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256183)
- [Yi-Fu Wu US](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/222fe091-3c48-42c9-8c29-a49579407053) — Software Engineer at NVIDIA
  [The Nemotron Ecosystem: Building the Third Generation — Nano, Super, and Ultra](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256810)
- [Yihao Wang](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/1cca8ab8-c85d-489b-b5ac-102654e96482) — Member of Technical Staff, RadixArk
  [SGLang-Diffusion: High Performance Serving Framework for Offline and Real-time Diffusion Models](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1258314)
- [Yiwei Song](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/e20baba8-3441-4bc5-be4a-f51df2e19e5d)
  [Keynote: AI Model Customization Platform](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1298069)
- [Yonghua Lin](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/ad720a72-1a77-4779-8155-f25d9ea8dc56) — Innovative AI and System Research Pioneer with 20+ Years of Expertise in Technological Advancement and Strategy Leadership
  [Portable PyTorch Across AI Accelerators: A Triton Operator Stack from Eager Mode to vLLM](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256909)
- [Yu Guangye](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/c3488095-867e-43ca-85d8-704cb71ace2f) — Intel, AI Framework engineer
  [PyTorch Generalization: A Journey Toward Write Once, Run Anywhere](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1256608)
- [Yueming Yuan](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/48d5c0e6-8003-47ed-b57e-143aa2bb3eb5) — Member of Technical Staff, RadixArk
  [Miles: Enterprise-facing Agentic RL Framework](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1258156)
- [Yun Zhao](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/3f378b4a-b718-4c9f-ab48-07c8fc1df1d1) — Senior Engineer, Huawei Co,. Ltd
  [HiFloat: Democratizing Ultra-Low Precision Training and Inference in PyTorch Ecosystems](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1206584)
- [Zach DeVito](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/6958d047-b51e-42a3-8089-7fa4ad55c8f1) — Meta
  [Monarch - Single-Machine DevX at Distributed Scale](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1255147)
- [Zachary Xi](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/50d5e9da-716d-4afb-9026-da0b1769ba05)
  [vLLM Roadmap](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1355211)
- [Zesheng Zong](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/c20c52c1-5413-485d-acc8-de4412276233) — Huawei, AI Framework Engineer
  [Relay and Reuse: The Dual Engine Behind PyTorch Out-of-Tree Release Readiness](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1253488), [Demo Theater: Agile Readiness: Leveraging CI Relay for PyTorch Workflows](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1361630)
- [Zhanqiu Hu](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/9a8cea08-5e46-4ff5-9dc2-c288d9d6d4a8) — Machine Learning Engineer, Red Hat
  [State-of-the-Art KV Transfer for Disaggregated LLM Serving in vLLM](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1255023)
- [Zhipeng Wang](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/be231dff-91a9-4bd9-8fba-2b8870f4a8d8)
  [Keynote: Scaling Large Frontier Models Training with DeepSpeed](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1314372)
- [Zhipeng Wang](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/5c383e29-ecda-4a70-85c3-a398e8072545)
  [Meet the Developers of DeepSpeed](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1364544)
- [Zhiyu Cheng US](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/a57abaca-3516-4c14-992d-f030968e556f) — Engineering Manager, Model Optimizer, NVIDIA
  [Fast, Small, Accurate: Optimizing Models for Inference using FP4 Quantization](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1253987)
- [Ziming Zhou](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/96d8e713-f9ce-4dbc-ac02-9548431d3420) — University of Michigan & ByteDance Seed
  [Before the Loss Spike: Training Alignment for Precise PyTorch Debugging at Scale](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1199488)
- [Zongwei Zhou](https://sessionize.com/api/v2/s2rq62nk/view/markdown/speaker/fc217fd0-6492-4f48-a27a-0c0c6f453fc3)
  [Keynote: Workload Fungibility in the Agentic Era: Bringing Native PyTorch to Cloud TPUs](https://sessionize.com/api/v2/s2rq62nk/view/markdown/session/1298062)