Hey, I'm Santosh 👋

I'm an AI Engineer Lead at Syngenta, based in Pune, India. By day, I design and ship LLM-powered products — retrieval-augmented generation, evaluation pipelines, and the MLOps plumbing that keeps it all running in production. By night (and most weekends), I write about it here.

The short version of how I got here

I started out as a full-stack engineer who genuinely enjoyed the messy parts — wiring up TypeScript backends, debugging React, figuring out why a deployment broke at 2 a.m. Over time, I kept getting pulled deeper into the stack: first analytics, then machine learning, and now generative AI and agentic systems.

Working across the full surface — from PyTorch and SageMaker to Node.js, AWS, and Databricks — taught me something most blog posts skip: a model is the easy part. The hard part is everything around it. The retrieval layer, the evaluation loop, the failure modes, the cost curve, the team that has to maintain it on a Tuesday morning six months from now. That's what I find interesting, and it's what I write about.

What I focus on

I'm less interested in "here's a cool new framework" and more interested in three things: How systems actually work end-to-end. Not the model in isolation — the full pipeline. Where the data comes from, how it's served, how it's evaluated, how failures surface, how the loop closes. Most of the value, and most of the bugs, live between the boxes on the architecture diagram. Why a technical choice matters to the business. Lakebase vs Lakehouse, batch vs streaming, RAG vs fine-tuning — these aren't just engineering preferences. They show up in latency the user feels, dollars the company spends, and the risk profile of the system. I try to make those tradeoffs visible. The technical details that separate a demo from production. Eval harnesses, observability for non-deterministic systems, drift, guardrails, the unglamorous work of making something reliable at scale.

What I write about

The topics I keep coming back to:

Generative AI & agents — agentic AI, Model Context Protocol, agent-to-agent communication, and what actually works in production. Data platforms — Lakebase, Lakehouse architecture, Delta Lake, Delta Live Tables, data fabric, OLTP for AI agents. Engineering craft — TypeScript, Node.js, NestJS, micro-frontends, system design. Honest takes — anti-patterns I've seen, mistakes I've made, and things I wish someone had told me earlier.

If a post can't help someone build, debug, or decide something a little better, I'd rather not publish it.

Let's connect

🌐 Personal site — santoshshinde.com 💻 GitHub — @santoshshinde2012 💼 LinkedIn — in/shindesantosh If something I write here helps you ship a project, avoid a pothole, or make a sharper call — drop me a note. That's what makes this worth doing.