Assess
Azure-native alternative to MLflow. Provides built-in MLflow compatibility, meaning any experiment tracking done in MLflow today can be pointed at an Azure ML backend without code changes. Higher operational overhead and cost than a self-hosted MLflow instance, but eliminates a separate managed service.
Assess as part of the Phase 2 ML pipeline maturity work. If the team grows and Azure ML's managed compute and endpoint features become compelling, the migration path is low-friction. See DEC-006.