Sanika Kotgire
AI & Data Engineer @ ZS | AWS Community Builder | Author | Public Speaker
Pune, India
Actions
I'm an AI Engineer & Data Engineer passionate about building intelligent, scalable, and cloud-native systems. I'm an AWS Community Builder, public speaker, technical writer, and Author of Soaring Brimstone's Flight. I actively organize and lead AI, cloud, and developer-focused technology conferences, workshops, and community events, helping connect engineers with emerging technologies and open-source innovation. I enjoy sharing practical lessons through speaking and community engagement.
Area of Expertise
Topics
Building Production-Ready GenAI Workflows on AWS
A GenAI prototype can be built in an afternoon. Making it reliable enough to survive production is a very different problem.
In this session, we’ll build and break down a serverless GenAI workflow on AWS using Amazon Bedrock, AWS Lambda, Amazon API Gateway, Amazon S3, and event-driven components. Rather than focusing only on getting an LLM to return the right answer, we’ll look at the engineering problems that appear once real users, failures, latency, retries, and costs enter the picture.
We’ll explore what happens when model calls time out, downstream components fail, requests are retried unexpectedly, or an AI workflow becomes too expensive to operate at scale. We’ll then walk through practical patterns for adding failure handling, observability, asynchronous processing, retries, and cost controls without turning the architecture into unnecessary complexity.
Attendees will leave with a practical blueprint for taking a GenAI application from “the demo works” to “I’d actually deploy this.
Code Mode vs. Tool Calls: When Should Your Agent Write Code Instead of Calling MCP Tools?
MCP made tool calling the default way agents interact with the world, but a growing pattern called "Code Mode" has agents write and execute code that calls MCP tools programmatically instead of invoking them one at a time.
This talk breaks down the real tradeoffs: when tool calling wins (simplicity, auditability, guardrails) and when Code Mode wins (composability, fewer round trips, handling large tool surfaces without blowing the context window). Using concrete examples, we'll look at where each pattern breaks down in practice and give developers a framework for deciding which to reach for as they design their own MCP-based agents.
Cutting LLM Cost with Semantic Caching on Valkey Search
You already run Valkey as a cache. Then a feature ships that calls an LLM, and you are suddenly paying seconds of latency and real money per request- behind a cache that never hits, because no two prompts are ever byte-identical.
Semantic caching fixes the key, not the cache. Embed the incoming prompt, find the nearest prompt you have already answered, and serve the stored response when similarity clears a threshold.
Valkey Search makes that a single FT.SEARCH- vector similarity, TAG scope and NUMERIC recency in one query, with no second datastore to run.
This session builds it in production terms:
1. the index schema, and the embed to search to threshold to serve loop
2. threshold tuning, where a false hit costs you more than a miss ever did
3. invalidation with per-field TTLs from 9.0, instead of blanket expiry
4. what it actually costs
We will walk hit rate, latency and spend before and after, including why the embedding step and not Valkey is usually the bottleneck.
You will leave able to cut LLM spend using the Valkey you already run.
Detecting and Handling Silent Failures in Agentic AI Workflows
Agentic AI workflows don't usually fail loudly, they fail silently. A tool call returns the wrong result but no error, an agent takes a plausible-but-incorrect path, or a multi-step chain quietly drifts off course without ever throwing an exception. In this lightning talk, I'll break down the common silent failure modes in agentic systems built on AWS (Bedrock Agents, multi-step tool orchestration), why traditional error handling and monitoring miss them, and practical detection strategies you can add today: output validation checkpoints, tool-call auditing, and lightweight evals that catch drift before it reaches users. The goal is a clear, actionable framework for building agentic workflows you can actually trust in production.
License and SBOM Gaps in AI Agent Tooling
As AI agents increasingly rely on the Model Context Protocol (MCP) to connect with tools and data sources, developers are pulling in third-party MCP servers at a rapid pace, many built by small teams with inconsistent licensing practices and little to no SBOM coverage. This creates a compliance blind spot that most organizations haven't caught up to yet: unclear license terms on community-built MCP servers, missing dependency manifests, and unvetted code running with elevated permissions inside agent workflows. In this talk, I'll walk through what this risk actually looks like in practice and share concrete steps teams can take today: auditing licenses properly, generating SBOMs for MCP servers, and putting basic governance checkpoints in place. The goal is simple: bring the same supply chain discipline we already apply to traditional open-source dependencies to this fast-moving new layer of AI tooling.
MCP as an Open Standard: Avoiding Vendor Lock-In in the Agent-Tool Interop Layer
The Model Context Protocol has rapidly become the default way LLMs connect to tools and data — but as adoption accelerates across vendors (Anthropic, OpenAI, Google, and dozens of platforms), the open-source community faces a familiar risk: fragmentation through vendor-specific extensions that quietly break interoperability. This talk looks at what's needed to keep MCP genuinely open as it scales SDK conformance across implementations, the role of open governance now that MCP sits under the Agentic AI Foundation, and where early cracks in cross-vendor compatibility are already showing up. We'll walk through concrete examples of interop failures between clients and servers, and what the ecosystem is doing (and should do) to prevent MCP from repeating the fragmentation seen in past "open" protocols.
Your DPU Doesn't Speak to My DPU: Fixing Networking's Tower of Babel
As networks evolve into dynamic, programmable platforms for AI-native and edge workloads, DPUs and IPUs have become critical infrastructure offloading networking, security, and data movement from the CPU. But every vendor has built its own stack, APIs, and tooling, creating fragmentation that slows adoption and locks operators into single-vendor ecosystems.
In this lightning talk, we'll break down why this fragmentation exists, and show how the open source community is solving it through a vendor-neutral, standards-based control plane for DPU/IPU silicon. No deep architecture dive just the problem, the open source fix, and how you can get involved.
Why AI Agents Forget: Engineering Memory Systems with Valkey
LLMs can write code, answer questions, and execute tasks, but they have one major limitation: they forget. Every new conversation starts from scratch unless we explicitly build systems that preserve context, state, and memory.
In this talk, we'll explore what it actually takes to give AI agents memory. Rather than focusing on prompts or models, we'll focus on the infrastructure layer that sits behind them. We'll examine common memory patterns used in agentic systems, including conversation history, session state, semantic caching, user preferences, and long-term knowledge retrieval.
Using Valkey as the foundation, we'll discuss design decisions, trade-offs, and operational challenges such as memory growth, retrieval latency, cache invalidation, and observability. We'll also look at how memory architectures impact both response quality and inference costs.
The goal is not to build another chatbot. The goal is to understand how stateful AI systems are engineered and why memory is becoming one of the most important infrastructure problems in modern AI.
DPUs/IPUs at the AI Inflection Point
As AI workloads reshape modern data center architecture, the humble network card has evolved into one of the most strategically important pieces of infrastructure in the stack. DPUs (Data Processing Units) and IPUs (Infrastructure Processing Units) now handle networking, storage, and security offload at scale freeing CPUs and GPUs to focus entirely on compute, and enabling the kind of east-west traffic and multi-tenant isolation that AI-scale data centers demand.
But this shift comes with a challenge: every vendor has built its own stack, APIs, and tooling, creating fragmentation that slows adoption and locks operators into single-vendor ecosystems.
This session explores how the Open Programmable Infrastructure (OPI) community is tackling this problem head-on building a vendor-neutral, standards-based control plane that works across silicon from Nvidia, Intel, AMD, Marvell, and others.
From Code to Creativity: Exploring the Role of Women in Tech Innovation
"From Code to Creativity: Exploring the Role of Women in Tech Innovation" is a talk that celebrates the remarkable contributions of women in the tech industry. It highlights how women are not only excelling in traditional coding roles but also bringing their unique perspectives and creative abilities to every field from cloud computing and Generative AI ,to cybersecurity and blockchain, there are no boundaries left! The talk will delve into specific examples of how women are leading innovation in these areas, from designing scalable cloud infrastructures to developing cutting-edge AI algorithms. By showcasing the diverse talents and expertise of women across the tech landscape, the talk aims to inspire and empower others to pursue careers in technology and drive positive change in the industry.
MCP-Powered AI Agents in Kubernetes
In this session, I explored how Model Context Protocol (MCP) enables seamless orchestration of AI agents within Kubernetes and demonstrated how Kubernetes can be integrated with LLMs through live, end-to-end demos. The session also covered hands-on prompt execution and agent workflows in a cloud-native environment, followed by a practical architecture walkthrough showcasing how intelligent AI workflows can be designed, deployed, and scaled for production.
Grafana Mimir: Scaling Prometheus Metrics like never before
In this session, I discussed the scaling limitations of Prometheus and how Grafana Mimir addresses them through its distributed, horizontally scalable architecture. I covered the core components of Grafana Mimir, its read and write flow, and the architectural patterns that enable it to efficiently handle large-scale metrics. The session concluded with a live technical demonstration of configuring Prometheus for remote write to Mimir, setting up Mimir as a Grafana data source, visualizing metrics through dashboards, and exploring the Mimir Admin UI along with its key configuration files.
The Hidden Cost Drivers in GCP Architecture
In this session, I explored the hidden cost drivers in Google Cloud Platform (GCP) architectures and the architectural decisions that often lead to unnecessary cloud spend. I covered key areas including data transfer and egress costs, persistent disk and storage optimization strategies, and the impact of idle compute resources. The session focused on practical techniques for identifying cost inefficiencies and designing cloud-native architectures that are both scalable and cost-effective.
Live open-source contribution and project demonstration
This session explored how GitHub has evolved beyond a code hosting platform into the foundation of modern software development and collaboration. Participants learned how to build efficient Git workflows, manage repositories effectively, and collaborate seamlessly across teams using industry best practices. The session also introduced GitHub Actions for automating CI/CD pipelines, demonstrated real-world open-source contribution workflows through a live project, and covered managing GitHub Organizations and GitHub Pages. By combining practical demonstrations with real-world development scenarios, attendees gained the skills needed to use GitHub confidently for both individual and collaborative software projects.
Evolution of modern software delivery and infrastructure
This session provided a comprehensive introduction to the modern DevOps ecosystem, covering the evolution of Cloud Computing, DevOps, and Platform Engineering and how they work together to enable scalable software delivery. Attendees explored the latest industry trends, learned the essential skills and tools expected from DevOps engineers, and received practical guidance on securing internships and entry-level roles. The session also highlighted impactful hands-on projects that help aspiring engineers build real-world experience, strengthen their portfolios, and stand out in today's competitive job market.
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