Case Study: Exa achieves sub-200ms entity search with Zilliz Cloud

Exa cuts neural search latency from seconds to under 200ms with Zilliz

Exa, a company building a search engine for AI agents, needed a high-performance vector database to power its new entity search product. The challenge was to support hybrid retrieval, complex metadata filtering, and frequent data updates while delivering millisecond-level latency, all without diverting engineering resources from their core product. They turned to Zilliz Cloud, the fully managed Milvus service from vendor Zilliz, to meet these demanding requirements.

By implementing Zilliz Cloud, Exa gained a solution with native hybrid search capabilities, combining dense and sparse vectors with metadata filtering in a single API call. The results were significant: neural search latency was reduced from seconds to under 200ms. The fully managed service also provided high reliability with near-zero operational incidents, freeing Exa's engineering team to focus on product development. Zilliz Cloud's performance and operational simplicity enabled Exa to scale its entity search product quickly and confidently.

Exa