Now

Updated September 1, 2026

Building

AI product work at Advance Local, on the newsroom AI innovation team. Two platforms: a self-service layer that hundreds of automated jobs run through, and a reusable system for turning raw data into published, embeddable stories. I also operate them, which this month has meant model migrations across a live fleet and cutting alert noise that was training people to ignore alerts.

Also Between Visits, where I am product and technical lead. It is a text-message bridge for families in pediatric developmental follow-up, who can see six specialists in one visit and then not return for a year. It runs on synthetic data while the pilot is earned. Most of my hours on it lately have gone to carrier compliance registration for the SMS layer, which is less paperwork than it sounds and more of a design constraint: what you are allowed to send, and what you have to prove about consent before you can send it.

Thinking about

Why the measurement problem in this field is really an operations problem: engines change their answer surfaces without notice, and every change silently breaks a collection layer. Whoever maintains collection best wins, regardless of who has the better dashboard.

Available for

Senior product and product engineering roles building AI systems: platforms, automation, evaluation, and measurement. Email me.