01Product ML
Recommendation, ranking, candidate retrieval, personalized discovery, and product scoring.
- Candidate generation and retrieval
- Learning-to-rank and product scoring
- Recommendation and personalized discovery
~12M
monthly users reached by personalized decision systems
See personalized product discovery ↗02Causal measurement
Experiment design, uplift modeling, holdouts, confidence intervals, and decision-oriented measurement.
- A/B and factorial experimentation
- Propensity, uplift, and treatment effects
- Persistent holdouts and confidence intervals
~US$1M
monthly incremental GMV measured against holdout
See causal ML for remarketing ↗03Reliable AI systems
Production serving, monitoring, MLOps, and rigorous offline evaluation for ML and applied LLMs.
- Batch and real-time model serving
- Monitoring, drift, and experiment tracking
- LLM evaluation and retrieval metrics
200+
pull requests reviewed across distributed engineering teams
See production ML serving ↗ PythonGoRSQLBigQueryRedisGCPAWSPyTorchTensorFlowLightGBMXGBoostscikit-learnOptuna