About
30+ years building models that make real decisions, at volume, in production.
My work started in credit, fraud and identity, where a model approves loans, stops fraudulent transactions or decides whether two records are the same person, and every error has a cost. That work taught a discipline that carries straight into today's AI systems: test on real data before launch, measure accuracy honestly, monitor after deployment, and fix what drifts.
Today I build AI agents, LLM applications and quantitative models, and I advise companies and investors on where AI will and won't pay off.
Scoring models that approve or decline at scale
Finding rare events in high-volume data
Matching records about the same person across messy sources
Response and propensity models that decide who to contact
Systematic models for financial markets
Models and agents deployed, monitored and maintained