Sakalya Mitra

In 2018, AlphaZero taught itself chess and took apart the best engines in the world. I'd grown up on chess — watching a machine learn like that is the moment I knew I'd do AI. In 2020 I chose an Integrated M.Tech in AI at VIT Bhopal: theory plus research, because I wanted to find things out, not just use tools.

I tried everything — full-stack, Android, web3, security — and none of it stuck. Then I wrote my first ML model in class, a linear regression predicting house prices, and it clicked. I went deep: machine learning, deep learning, 10 peer-reviewed papers. Then ChatGPT landed and I had to understand it from the inside out.

So I chased the real thing — an NLP pipeline at NinjaStudy (now YC-backed Speakify), data at scale at Scaler, then FutureSmart AI, where I led three projects. The biggest, Glamira, went live across 80+ countries for 10,000+ users a day. I joined Sanscritic for the next challenge: scalable systems that bring founders' and domain experts' visions to life.

The throughline: I care whether something is actually right, not just whether it looks right — the same instinct behind solving Rubik's cubes, photographing while I travel, and writing to make hard things simple. Next: build AI that solves real problems at scale, while going deeper into the anatomy of LLMs and everything changing in them, week to week.