What is a model registry, and why is model versioning important?

A model registry is a central catalog that stores trained model artifacts along with their metadata: version, training data reference, hyperparameters, evaluation metrics, lineage, and stage (e.g., staging, production, archived). Tools like MLflow Model Registry, SageMaker, and Vertex AI provide this. Versioning matters because production models change frequently; you need to know exactly which model version served a given prediction for debugging, auditing, and compliance. It enables safe promotion workflows (approval gates before a model goes to production), fast rollback to a known-good version if a new one degrades, reproducibility by linking a model to its exact code and data, and A/B comparison across versions. Without a registry, teams lose track of what is deployed and cannot reliably reproduce or revert models.