Physicist by training, AI engineer by trade. I build ML systems and write about their strange, poorly understood failure modes, the stuff that looks fine in a demo and breaks in production.
- ✍️ I write at my newsletter on the failure modes and architectures of modern ML
- 🔬 Currently building a SaaS and running reproducible ML experiments at Science_ML
Open-source contributions
- DashAI, Chile's public AI framework: contributed a BART regression model and a Matthews correlation coefficient metric (merged into develop)
- auto-bayesian, Santander AI Lab: fix for a silent NaN-handling bug in one-to-many aggregates (PR open)