Category
Keep your models healthy in production. Detect drift, version datasets, and automate retraining to close the MLOps loop.
The final step in the MLOps lifecycle — operating and monitoring models in production.
Learn the three types of drift (feature, label, prediction), implement KS tests and PSI scores for detection, generate automated reports with Evidently AI, version datasets with DVC, and wire up Airflow to automatically trigger retraining when drift is detected.
Operations & MonitoringSet up DVC as "Git for Data" — learn how pointer files work, configure S3 remote storage, and follow a step-by-step walkthrough to version your ML datasets so every model can be traced back to the exact data it was trained on.
Completed all the guides? You now understand the full MLOps lifecycle from data ingestion to production monitoring and dataset versioning with DVC. Go back to the beginning to review, or explore all guides.