Y

MLOps · Advanced · 6 h

Monitoring, drift and retraining

Detect data drift and falling performance, and decide when and how to retrain.

Why employers care: Models decay silently. Senior ML roles own the monitoring that catches it before the business does.

In the roadmaps for: Data Scientist, AI/ML Engineer

Lessons, in order

  1. When a good model goes bad: drift and decayA model with 0.76 ROC AUC falls to 0.53 on later data. Detect it with drift metrics, and plan retraining.4 min