Module 5 · MLOps: Deploying Machine Learning & Deep Learning Models
Monitoring and Retraining: lessons
Drift detection, performance with delayed labels, retraining strategies, and safe rollouts.
3 lessons · 48 min
What this module teaches, what it asks of you first and how you practise it are on its syllabus page.
Open the syllabusDetecting Data Drift
18 minPSI with quantile bins, the KS test, why p-values are always "significant" on big data, and how to calibrate thresholds so the alarm does not fire every morning.
Included with the track Simulation insideRetraining Strategies
15 minWhen to retrain — on a schedule, on drift, or on performance — on which data window, and why a trigger is not permission to release.
Included with the track Simulation insideShadow and Canary Deployments
15 minRunning a new model beside the old one — with no consequences (shadow), then on a small slice of traffic (canary) — and computing how much traffic it takes before the result can be trusted.
Included with the track Simulation inside