Accelerate ML feature pipelines with new capabilities in Amazon SageMaker Feature Store

This article announces three new capabilities in Amazon SageMaker Feature Store (SageMaker Python SDK v3.8.0) that address operational challenges in production ML platforms.

• Native AWS Lake Formation integration for automatic column, row, and cell-level access control
• Additional Apache Iceberg table properties to control metadata retention and prevent cost overruns
• Full Feature Store support in modernized SageMaker Python SDK v3 with modular architecture
• Lake Formation automatically registers S3 locations and manages access without manual setup
• Iceberg properties like metadata deletion and snapshot limits prevent metadata accumulation
• One customer reduced 50TB metadata growth by enabling metadata cleanup properties
• Online store unaffected; Lake Formation applies only to offline store
• Properties can be set at feature group creation or applied to existing groups

These enhancements automate access control and cost management for ML feature pipelines, eliminating manual configuration overhead and reducing unexpected storage costs.