Module MD10
Model lifecycle and MLOps
Manage datasets, models, drift, and retraining as controlled, reviewable changes.
Why this module
Manage datasets, models, drift, and retraining as controlled, reviewable changes.
What you produce
A small model lifecycle pipeline. Assess the output against the unit outcomes, explanation, safe handling, and independent transfer. Do not require public sharing or use of real sensitive records.
Chapters and units
MD10-C01: Traceable experiments and artifacts
Reproduce a training run with traceable inputs; Link a model version to data, code, and evaluation
- MD10-U01 Datasets, features, and experiment records Reproduce a training run with traceable inputs
- MD10-U02 Model registries and artifact lineage Link a model version to data, code, and evaluation
MD10-C02: Serving and change detection
Choose a scoring pattern and monitor outcomes; Distinguish distribution change from measured quality loss
- MD10-U03 Serving, batch scoring, and monitoring Choose a scoring pattern and monitor outcomes
- MD10-U04 Drift, labels, and retraining decisions Distinguish distribution change from measured quality loss
MD10-C03: Controlled releases and ownership
Reject an inferior candidate and restore the previous model; Assign owners for data, model, incidents, and change
- MD10-U05 Release gates and rollback Reject an inferior candidate and restore the previous model
- MD10-U06 Lifecycle handover and review Assign owners for data, model, incidents, and change