Engineer and product leader. I build systems in code, wood, and fire.

I stepped from engineering into product management and never stopped building. I joined Google in 2014 wanting to take a product from zero to one. Twelve years on, that remains the work. Product isn't a career track to me — it's what I would be doing anyway.

It started with a table of compensation data and hours spent writing manual filters. Get every filter right, and nobody notices. Get one wrong, and you've exposed what people are paid. I wasn't trying to secure data in the abstract; I was trying to protect myself and the engineers around me. So I built the engine that writes the filters: declare the rule once in plain business logic, compile it to whatever each downstream system speaks. Define once, enforce everywhere.

That exact problem kept recurring under different names. First with data contracts — what a dataset guarantees, declared once by its owner rather than guessed downstream. Then a knowledge graph, linking disparate domains that had never shared an identity. Now AI agents, where the rule lives inside a system prompt and quietly fails under drift. As the scope grew, the medium shifted: I stopped building every component myself and began building the teams that did, eventually handing my first platform to a PM I hired.

Off the clock, I build the same systems at a scale of one: a multi-agent harness I run locally, and a small policy engine to govern it. Outside the terminal, I mill rough-sawn hardwoods, pour epoxy river tables, and manage charcoal heat at the tandoor. On the bench, you build a jig because human attention degrades by the fortieth cut, and four-quarter walnut isn't forgiving. A fail-closed policy boundary is the exact same fixture. Respect the grain of the system. Finish honestly.

Most of what I think is under Writing. Most of what I make is under Projects. If you're working on the same problems, write to me: ankit@ankitskgoyal.com.