Selected work · 2026
One system, taken end to end: rigor that ships.
A deployed ML system — scraping and deduplication, a leak-free evaluation, drift monitoring, and a serving API. The same discipline throughout: work out what the problem actually needs, then build only that.
01Deployment · MLOpsLIVE DEMO2026
Car Price Prediction & MLOps
End-to-end ML system on LightGBM with TF-IDF+SVD text features — scraping, dedup, leak-free 5-fold evaluation, drift monitoring, SHAP explainability, and a FastAPI serving layer. The complete production cycle, not just a notebook.
LightGBMTF-IDF+SVDFastAPIDuckDBRailwayNext.js
6.49%
out-of-fold MAPE
0.975
cross-validated R²
29,988
listings modelled
More end-to-end projects are in progress — the code lives on GitHub in the meantime.
GitHub ↗