Model Registry

A model registry is a centralized repository for storing, versioning, and managing trained machine learning models and their associated metadata — including training parameters, evaluation metrics, data lineage, and deployment history. It serves as the authoritative source of record for which model versions exist, what their performance characteristics are, and where they are deployed.

In enterprise AI operations, a model registry is foundational infrastructure for governance, reproducibility, and operational control. When multiple teams are building and deploying models across a commerce platform — for search ranking, personalization, fraud detection, and forecasting — a registry ensures that deployments can be audited, rolled back, and traced to specific training runs. It also enables promotion workflows (moving a model from staging to production after approval), A/B testing governance, and compliance documentation. Major MLOps platforms such as MLflow, SageMaker, and Vertex AI include model registries as core components.