PRODUCTION SCENARIO
A bank retrains a credit-limit model each month. Risk officers must approve each version before it is deployed. Auditors need to trace which version was in production on a given date, along with the training metrics for that version.
Which solution will meet these requirements?
Answering here is anonymous. Nothing is saved unless you sign in.
Show answer and explanation
Answer: Register each trained model as a version in a Model Registry model group.
Model Registry does all three jobs: it adds each trained model to a model group as a new version, associates training metrics with it, manages approval status, and records lineage for traceability. Experiments captures the metrics but tracks training runs rather than the deployed artifact, and has no approval state. Model Monitor watches quality after deployment, too late for a pre-deployment gate. A Model Card documents how a model was built and is not the deployment gate: it does not version the model or record which version was serving traffic.