A production-grade AIOps framework focused on model integrity and autonomous reliability. Features a LightGBM-driven multiclass inference engine via FastMCP, validated against data leakage through NMI analysis. Includes real-time Population Stability Index (PSI) drift monitoring and a closed-loop human-in-the-loop feedback system for sustainable ML

This project builds a production-grade ML pipeline to classify Near-Earth Objects (NEOs) as hazardous or non-hazardous. It automates data ingestion, preprocessing, model training, monitoring, and drift detection using GitHub Actions, PostgreSQL, MLflow, DAGsHub, and Grafana.

Transform CI/CD failure logs into real-time automated anomaly detection and root-cause classification using supervised machine learning.