Homagni Saha
Senior Applied Machine Learning Scientist
Actions
Homagni Saha is a machine learning and computer vision expert. He earned his PhD degree from Iowa State University, where he focused on building optimized machine learning systems for vision perception and robotics applications. With 3+ years of experience, he has led the design of computer vision solutions for off-highway vehicles at Danfoss and developed multiagent algorithms for aerial vehicles at Honeywell. His research spans diverse areas including vision-language navigation, multimodal information fusion, 2D to 3D machine learning, efficient machine learning techniques, and multi-agent robotics. He is the author of over 20 research papers with 250+ citations, holds 2 filed patents, and has served as a reviewer for several top-ranked journals and conferences. Homagni received his concurrent master's degrees in mechanical engineering and computer science from Iowa State University in 2018 and 2020 respectively, and graduated with a PhD in 2021. Outside of his professional pursuits, he is passionate about 3D printing and enjoys learning to play new musical instruments. During his leisure time, he engages in experimenting with new 3D printing models and creating 3D reconstructions using photogrammetry.
Demos: Revolutionizing Grainger with AI & MLOps
Demo 1: GitOps for Model Serving: Leveraging ArgoCD, Flagger, and Flux Presented By Victoria Reese
The Model Serving Platform leverages numerous Kubernetes-native GitOps tools for production support of model serving endpoints including Flagger and Flux for progressive delivery and ArgoCD for continuous deployment. This session will provide an overview of these tools and why the MSP team decided to leverage them.
Demo 2: Harnessing the power of AI/ML with Snowflake Cortex Presented By Chirag Gupta & Mayra Vazquez
This presentation will demonstrate how to use Streamlit and Snowflake Cortex to create user-friendly data science applications directly within Snowflake. We will explore building interactive UIs with Streamlit and integrating AI/ML capabilities using Snowflake Cortex. By keeping data within Snowflake, we can enhance data insights and automate predictions for Grainger teams.
Demo 3: Multi camera 3D capture of objects with challenging textures Presented By Engin Anil & Homagni Saha
Grainger aims to improve product search by using 3D models instead of relying solely on text or images. Creating accurate 3D models of products is challenging due to limitations of current 3D scanning technology. This demo presents a new approach to capture detailed 3D information using a single camera and advanced image processing techniques, offering a more efficient and cost-effective solution for creating 3D product models.
Multi camera 3D capture of objects with challenging textures
Being able to retrieve similar types of products a user is searching for is an important core technology for Grainger. Capturing and utilizing exact 3D geometry of the product SKUs (including colors and textures) has significant potential in improving search and matching which currently mostly use matching based on textual descriptions and sometimes images (visual search). Availability of large-scale geometry for 3D matching is limited by current 3D scanning technologies which either require expensive 3D sensors and setups or significant scanning times and processing powers compared to the easier route of taking multiview images for matching. This demo explores a solution to obtain pixel level accurate 3D scans of small texture less objects using single camera setup while reducing scan times and processing costs by combining concepts of projective geometry with recent advances in neural network based semantic segmentation techniques.
Homagni Saha
Senior Applied Machine Learning Scientist
Actions
Please note that Sessionize is not responsible for the accuracy or validity of the data provided by speakers. If you suspect this profile to be fake or spam, please let us know.
Jump to top