I build AI products and carry them through the part nobody talks about: the security review, the bureaucratic hurdles, the legacy systems designed for other purposes. Inside the companies where pilots go to die: $8B enterprises with decades of data and no software culture.

Data scientist, software engineer, builder of products and companies. Harvard CS. Today: enterprise AI for the built world, from ~$1B receivables forecasting to injury-risk models on live job sites. Most of what's interesting about me doesn't fit on a resume.
Models and data platforms on one side, security reviews and skeptical rooms on the other. Enterprise AI where being wrong is expensive: billions in receivables, live construction sites, people who have to trust it.
I write about things that interest me and may interest you: AI, Art, Philosophy, Language, Science. Why a model lands in one company and dies in another, and other patterns I keep finding in places that shouldn't have much in common.