Which MLOps control BEST ensures that only validated model versions are deployed to the production environment?
Select an answer to reveal the explanation.
Short Explanation
You need a control that actually blocks bad deployments, so use a model registry with promotion workflows that require validation sign-off before production tagging. That creates an enforced gate, not just a record or a later alert.
Full Explanation
The model registry with promotion workflows that require validation sign-off before a model version can be tagged for production deployment is correct because it creates a technical gate that prevents unvalidated model versions from reaching production. By making validation sign-off a prerequisite for the production tag, the control is automated and cannot be bypassed by individual engineers. Maintaining a git repository with version control for all model training scripts maintains code history but does not enforce validation before deployment. Conducting daily performance monitoring of production models and alerting on metric degradation detects post-deployment problems but does not prevent unvalidated models from being deployed. Documenting all hyperparameter configurations in a model development log is documentation, not a deployment gate.