I build products and get people to use them.
I run Soda Labs, where I validate, build and launch new bets for the company, most recently Sodabase. I also shape positioning and messaging for Soda’s launches and campaigns.
Before that, I led developer advocacy at NannyML and worked as a data scientist. I studied physics, with detours through electrical engineering and industrial design.
I’m writing The Little Book of ML Metrics (1K+ GitHub stars, 127 pre-orders), which goes to print in December 2026.
Projects
Finds and fixes the Supabase tables your agent broke.
One permissioned access layer for agents and teammates.
An open-source reference book on how to evaluate ML models.
Soda Cloud as a command line for humans and agents.
Positioning, messaging and first design of Soda’s website.
The site for Soda’s “Data Quality is Solved” launch event.
Customizable data quality dashboards. I was the PM.
292 data science flashcards, improved daily by an AI agent.
Realistically dirty synthetic data for testing pipelines.
A study of how ML models degrade over time.
Handwritten snail-season signs from Portuguese cafés.
Photos of that sign from bars and restaurants across New York.
Every side project I’ve started and later let go.
Writing
When agents verify claims, the product does the convincing.
Don’t add more unintentional side salads to the world.
Not just interested. Not just curious. Obsessed.
Control isn’t the goal. It’s the lesson.
