Feast now supports Ray and KubeRay, which means you can run your feature engineering and embedding generation jobs distributed across a Ray cluster.

You can define a Feast transformation (like text → embeddings), and Ray handles the parallelization behind the scenes. Works locally for dev, or on Kubernetes with KubeRay for serious scale.

Process millions of docs in parallel

Store embeddings directly in Feast’s online/offline stores

Query them back for RAG or feature retrieval

All open source