A model that passed every test can still fail in production. The world changes, the incoming data stops looking like the training data, and accuracy slips without any error being raised. Monitoring is how you notice this early, by comparing live data against a reference set and watching for data drift and falling model quality.
You'll learn to monitor models with Evidently, an open-source library that generates drift reports, and then see how to automate the whole loop: scheduled drift checks in Airflow, alerts when too many columns move, and retraining triggered automatically.
Governance is the other half: knowing which model version is live, who approved it, what data it was trained on, and how it was documented. Practices like model registries and model cards make ML systems accountable, and they build on the tracking and registry work from earlier modules. This is the final module, so next up are the hands-on projects.
A walkthrough of Evidently AI, a free and open-source Python library for monitoring machine learning models.
A step-by-step tutorial on generating monitoring reports with Evidently AI to keep an eye on model performance.
Build an end-to-end drift monitor that checks each new batch against the training reference, sends a Slack alert, and triggers retraining.