About

I was the kid who thought in algorithms. In college, competitive programming was my playground and identity. The rush of AC, the elegance of an O(n log n) solution, the quiet joy of shaving constant factors — those were my dopamine loops. I didn’t just solve problems. I collected them. Codeforces handle, ranked contests, hackerrank badges — you name it.

But contests don’t run in production. After graduation, I joined Amazon. Suddenly, the problems weren’t well-posed anymore. They were messy, ambiguous, and had blast radius. I learned to care about p99, retries, idempotency, and the art of saying “no” to the wrong abstractions. Building distributed systems at scale is humbling — your naive assumptions meet millions of users.

Somewhere along the way, I fell in love with a different kind of problem: understanding ideas at the frontier. Reading a paper feels like entering someone’s mind. You can’t really learn until I reproduce it — line by line, bug by bug, until the curves match (or don’t). Reproducing ML papers is slow, frustrating work. Half the time, the numbers don’t budge. But when they do — it’s magic. You don’t just believe the result; you understand its bones.

This site is my lab notebook. I write to think, to remember, and to connect threads across systems, theory, and practice. I’m still figuring things out — one experiment, one post, one late night at a time.