Data scientist. I make machine learning earn its keep in real products.
For over a decade I’ve built models that price things, value things, and predict what people will do next — and kept them accurate, explainable, and trusted long after launch.
What I do
Production ML
Forecasting, valuation, and behavioral prediction models, owned end to end — from the data pipeline through retraining, monitoring, and the unglamorous work that keeps a model alive in year three.
Models people trust
Explainability and validation as first-class work: SHAP-based explanations and evidence a non-technical stakeholder can actually read, because a model nobody trusts is a model nobody uses.
Analysis that decides things
Segmentation, clustering, and ad-hoc investigations that turn a vague business question into a scoped answer — plus the dashboards that let people explore it themselves.
Good models earn their keep after the notebook closes.
Off the clock I build small software tools for myself, and I speak Cantonese, Mandarin, and English — plus just enough Japanese to order dinner and mostly get what I asked for.
Have a data problem you’re turning over?
I like other people’s puzzles — a model that won’t behave, a “do we even have the data for this” question, a side project that needs a sanity check. If it’s interesting, I’m in.