|
I'm an AI Lead Engineer at Syngenta, based in Pune. I design and ship LLM-powered products — RAG pipelines, evaluation loops, and the MLOps plumbing that keeps them running in production. My focus is AI/ML — powered by deep full-stack engineering roots. That combination lets me bridge the gap between a working model and a product people can actually rely on. The model is the easy part. The hard part is everything around it — retrieval, evals, failure modes, cost, and the team that maintains it six months from now. That's what I write about. |
I'd rather show working systems than list skills. Everything here runs and is inspectable:
| Working system | What it demonstrates | Link |
|---|---|---|
| FrameSleuth | A shipped, local-first AI product — video → evidence-cited context bundles for coding agents, MCP-ready | framesleuth.com |
| ask-santosh | Retrieval-augmented Q&A over my writing, with a reproducible DeepEval suite (faithfulness · answer-relevancy · contextual-relevancy) gating answer quality in CI | repo |
| Multi-agent on Databricks | Production multi-agent orchestration, beyond Genie code | repo · write-up |
| This profile | The summary just below is written by a GitHub Actions + GitHub Models workflow that reads my own recent activity — an agentic pipeline, not a static line | workflow |
Lately: building FrameSleuth (video → shippable code for agents), shipping ask-santosh (RAG over my writing with a DeepEval quality gate), and multi-agent + AI-consumption-plane work on Databricks — with a steady stream of write-ups on MCP and agentic systems.
Auto-generated from my recent GitHub activity by a GitHub Actions + GitHub Models workflow.
FrameSleuth is a local-first AI system that converts screen recordings into structured, evidence-cited context bundles for coding agents. Record a bug or a feature demo — it reads every frame, transcribes the narration, and captures console and network activity, then hands your agent repro steps, error evidence, and code candidates it can act on.
- End-to-end systems — not the model in isolation, the full pipeline. Most of the value (and most of the bugs) live between the boxes on the architecture diagram.
- Tradeoffs that matter to the business — Lakebase vs Lakehouse, batch vs streaming, RAG vs fine-tuning. These show up in latency, cost, and risk — not just engineering preference.
- The unglamorous production work — eval harnesses, observability for non-deterministic systems, drift, and guardrails. The stuff that separates a demo from something you can trust on a Tuesday morning.
A decade of building — from data pipelines and full-stack apps to production AI, with a habit of sharing what I learn along the way.
%%{init: {'theme':'base','themeVariables':{'fontFamily':'Fira Code, monospace','taskBkgColor':'#1F6FEB','taskBorderColor':'#2F81F7','activeTaskBkgColor':'#2F81F7','activeTaskBorderColor':'#58A6FF','doneTaskBkgColor':'#8B949E','doneTaskBorderColor':'#6E7681','critBkgColor':'#D97706','critBorderColor':'#B45309','critTextColor':'#ffffff','milestoneBkgColor':'#238636','milestoneBorderColor':'#2EA043','taskTextColor':'#ffffff','taskTextDarkColor':'#ffffff','taskTextLightColor':'#ffffff','taskTextOutsideColor':'#8B949E','gridColor':'#30363D','todayLineColor':'#F85149','titleColor':'#8B949E','sectionBkgColor':'#2F81F714','altSectionBkgColor':'#2F81F70A'}}}%%
gantt
title A decade of building — 2014 to today
dateFormat YYYY-MM-DD
axisFormat %Y
tickInterval 1year
section Craft
Data engineering (pipelines, Spark, Databricks) :active, de, 2014-01-01, 2026-07-01
Full-stack engineering (TypeScript, Node, React, AWS) :done, fs, 2014-01-01, 2021-06-01
Architecture and platform engineering :active, arch, 2018-01-01, 2026-07-01
section AI / ML
ML and data science :active, ml, 2021-06-01, 2026-07-01
LLM, RAG and agentic systems :active, llm, 2023-06-01, 2026-07-01
MLOps, evals and guardrails :active, ops, 2023-06-01, 2026-07-01
section Building in public
Answering on Stack Overflow :crit, so, 2015-01-01, 2026-07-01
Open source on GitHub :crit, gh, 2016-01-01, 2026-07-01
Writing on Medium :crit, med, 2019-01-01, 2026-07-01
section Milestones
AI Lead Engineer at Syngenta :active, syn, 2024-05-01, 2026-07-01
FrameSleuth launch :milestone, fsl, 2025-09-01, 0d
%%{init: {'theme':'base','themeVariables':{'pie1':'#2F81F7','pie2':'#238636','pie3':'#8B949E','pie4':'#D97706','pieStrokeColor':'#ffffff','pieStrokeWidth':'2px','pieTitleTextSize':'18px','pieSectionTextSize':'13px'}}}%%
pie showData
title Focus areas, right now
"AI agents, LLM and MCP systems" : 40
"Data and AI on Databricks" : 25
"Full-stack product engineering" : 20
"Technical writing" : 15
Not every problem lives in the same place. I map what I build to the Cynefin domains — because the right approach for a known CRUD API is the wrong approach for a non-deterministic agent.
AI, Agents & Products
| Project | What it is |
|---|---|
| FrameSleuth | Local-first AI that turns video into evidence-cited context bundles for coding agents. MCP-ready. |
| multi-agent-sales-ops-tpch-databricks | Beyond Genie code — orchestrating production multi-agent systems on Databricks. Write-up → |
| ai-consumption-plane | A hands-on build of the AI Consumption Plane on Databricks. |
| crop-disease-prediction | Crop-disease image classification — applied ML for agriculture. |
Data Engineering & Platform
| Project | What it is |
|---|---|
| medallion-architecture-databrics | Medallion Architecture — principles and a practical Databricks exploration. Read → |
| dataset-atlas | A map-first way to discover and download datasets — Region → Domain → Get. Live demo → |
| node-boilerplate — 460 stars | Production-ready Node.js + TypeScript skeleton for microservices — ESLint, Prettier, Husky, CI wired in. |
- The AI Consumption Plane on Data-bricks: A Hands-On Build
- The Medallion Architecture Reconsidered: What It Solved, and Where It’s Cracking
- Building a Production MCP Server That Handles Real API Complexity (Part 2)
- How Production Teams Are Designing MCP Systems That Actually Scale (Part 1)
- Building an AI Video Agent: Turning Recordings into Tests (Part 6)
- Building an AI Video Agent: One Engine, Many Bundles (Part 5)
- Building an AI Video Agent: The Bundle Is the Contract (Part 4)
- Building an AI Video Agent: Reading the Screen (Part 3)
- Building an AI Video Agent: From WebM to Keyframes (Part 2)
- Building a Local AI Video Agent: Architecture, Patterns, and the Hard Problems (Part 1)
More on Medium →
Open to conversations about production LLM systems, MLOps, and agent tooling — reach me on LinkedIn.