Selected impact
Machine learning that reaches production and moves the metric.
Selected professional case studies spanning ranking, causal ML, high-scale serving, and LLM evaluation. Public details are limited to non-confidential, high-level system design and already disclosed results.
Personalized Product Discovery & Ranking
Replacing a static affiliate landing page with a continuous, personalized discovery feed, measured through a multi-country controlled experiment.
Causal ML Remarketing Engine
An end-to-end decision engine for audience, timing, and content, evaluated against persistent holdouts across eight countries.
Production ML Serving Architecture
A hybrid batch and real-time architecture turning model scores, live behavior, and campaign rules into low-latency personalization decisions.
Applied LLM Evaluation for Recommendations
An offline evaluation methodology for retrieval and LLM-assisted recommendation systems, built around ground truth and uncertainty.