Vertex AI Feature Store is a feature management service in Vertex AI designed for efficient handling of machine learning (ML) features. It acts as a centralized repository, streamlining the creation, maintenance, sharing, and serving of ML features. There are two versions available:
- Vertex AI Feature Store: Utilizes a modern approach, serving feature data directly from a BigQuery source. It optimizes ultra-low latency serving, supports embeddings, and facilitates vector similarity searches.
- Vertex AI Feature Store (Legacy): Provides a centralized repository with both online and offline stores. It allows batch or stream import of feature data, online serving for predictions, but lacks certain capabilities like embeddings management.
Users can choose the version that suits their needs, with the newer Vertex AI Feature Store offering enhanced capabilities, including Dataplex integration for feature metadata tracking and advanced serving options.