Churn Signals
A gradient-boosted churn model with SHAP-based explanations, served behind a small FastAPI service and monitored for drift.
- Python
- XGBoost
- SHAP
- FastAPI
Selected work
A curated set of things I've built — models shipped, tools written, and ideas worth keeping. Each links out to the code.
A gradient-boosted churn model with SHAP-based explanations, served behind a small FastAPI service and monitored for drift.
Retrieval-augmented search over a personal library of ML papers — sentence-transformer embeddings, FAISS index, and a re-ranking step.
A small toolkit for probabilistic time-series forecasting with backtesting, prediction intervals, and clean diagnostic plots.
Notebooks and utilities for uplift modeling and treatment-effect estimation — a place to reason carefully about interventions.