Things I've built, tested, and written down.
Yukta
A health app built as an agent harness experiment. The code is public, but the write-up is about product judgment: what to automate, what to trust, and what to ship.
Read →PM coding tech stack
How I use AI coding tools to build functioning prototypes as a PM: the workflow, the guardrails, and the tradeoffs behind Yukta.
Read →My PM productivity system
How I decide what to automate, what to leave manual, and how I build the workflows around my own product work.
Read →Beyond evals: harness engineering for PMs
Why PMs working on AI products need to understand more of the technical system: harnesses, architecture, review loops, and where product judgment actually shows up.
Read →A bit about me.
How I work
I like problems that stay fuzzy at the start and get expensive if you guess wrong. I turn "AI could do this" into something a team can reason about: scope, tradeoffs, what to trust, what to verify. I care less about how many specs we write and more about what we are actually betting on.
Right now
I'm building in public: product experiments, agent workflows, my PM operating system, and essays on AI product thinking. This site holds the parts a resume cannot.
Off the clock
I read more than I post. I like hard problems. I am wary of ideas that only sound good in a deck.