MLOps · 2025
Predictive Maintenance Lab
Listening to machines before they fail.
The project
End-to-end sensor time-series ML system: sliding-window DSP features, XGBoost RUL (RMSE 16.80 on NASA CMAPSS) and bearing-fault detection (PR-AUC 0.9998 on CWRU), served via FastAPI with MLflow tracking, Docker, and CI/CD on Azure Container Apps.
At a glance
RMSE 16.80 · PR-AUC 0.9998 — NASA C-MAPSS RUL · CWRU fault detection.
Built with
Time-Series ML · MLOps · FastAPI · Docker