Model registries manage the lifecycle of trained models.
Features:
- Version control for models
- Stage transitions (staging → production)
- Metadata (metrics, lineage, owner)
- Access control
Workflow:
Train model, log to experiment tracker
Register best model in registry
Promote through stages (dev → staging → prod)
Deployment system pulls from registry
Tools: MLflow Model Registry, SageMaker Model Registry, Vertex AI
Interview tip: Know how models flow from training to production.