via “mlflow-based model training, versioning, and experiment tracking”
Unified analytics and AI platform — lakehouse, MLflow, Model Serving, Mosaic AI, Unity Catalog.
Unique: Databricks provides MLflow as a native, integrated experiment tracking and model registry system that stores all metadata and artifacts in the lakehouse, enabling tight coupling between training data versions (via Delta Lake time-travel) and model versions. Unlike standalone MLflow servers, Databricks MLflow is fully managed and integrated with the data platform, eliminating separate infrastructure.
vs others: More integrated than standalone MLflow (no separate server to manage), more comprehensive than Weights & Biases for teams already on Databricks (no additional SaaS cost), and provides better data lineage than SageMaker Experiments because models are versioned alongside the data they were trained on.