Which recommendation engine supports low-latency search for large catalogs with sparse user data?

For large catalogs with sparse user data, the best fit is usually a two-stage recommendation engine with an ANN-based candidate retrieval layer—for example:

• Amazon Personalize
• Google Vertex AI Search / Recommendations
• Milvus / Pinecone / Elasticsearch vector search used as the retrieval component in a recommender stack

If you want a single recommendation engine known for low-latency search at scale, Amazon Personalize is a strong answer because it’s designed for:

• large item catalogs
• implicit feedback / sparse interactions
• real-time recommendations with low latency

If you meant an open-source/search-backed option, then vector search engines like Milvus or Elasticsearch are commonly used to support low-latency candidate retrieval for recommender systems.

If you want, I can recommend the best option by use case: cloud-managed, open-source, or self-hosted.