I build the infrastructure AI runs on.

At DRDO I ran design-of-experiments on cutting forces across different metals. Vary the inputs, measure what comes out, let the numbers settle an argument that opinion couldn't.

I didn't think of it as data work. It just seemed like the most honest way to arrive at an answer. That instinct never left.

I raised an MSME grant to turn discarded banner material into eco-friendly bags. Later I co-founded Clapconcepts, a gifting company in Hyderabad, and Sketchmen, a digital marketing shop.

Different domain, identical instinct. What's the right product mix? Which channel actually converts? Where is the waste? I was running optimization problems before I had the vocabulary for them.

Overstock, then Backcountry, then Rakuten Rewards. Pulling dominant colors out of 100,000 product images to stop six-figure ad losses. Proving with Markov attribution that one channel was under-credited by 40%, which unlocked a $5M budget case.

By the end of it I'd stopped being the person who built the model and become the person who built what the models ran on.

I founded MasterControl's MLOps platform engineering practice from nothing, then architected GRID — a cloud-agnostic inference runtime that became the enterprise framework for every LLM and ML system in production.

Now I direct the whole function: strategy, architecture, and FinOps for global AI infrastructure, with a team across the US, India, Turkey, and Ukraine. Regulated life sciences means none of it ships unless it's observable, governed, and provable.

After a decade in the US I moved home. I sit on the program committee for IEEE ICA 2026, mentor and judge a 75-team cohort at T-Hub's FEG Innovation Hackathon, and advise Applied Biosensors on their GenAI work.

The part I enjoy most is the 1:1s — someone two steps behind me on a path I've already walked. That's what the booking links below are for.