Case Study: Cosmos achieves editorial-grade visual search with Qdrant Cloud

Cosmos cuts search latency to 200–500 ms with Qdrant

Cosmos, a visual search app for creative professionals, needed to power fast, multimodal search—including text, color, and hybrid queries—for millions of users. Their early prototype using Postgres and pgvector raised scalability and performance concerns. They required a managed vector search engine from a vendor like Qdrant that could handle multiple embedding types and rich metadata filtering without their team managing complex infrastructure.

The solution was implemented using Qdrant Cloud. Qdrant's named vectors allowed Cosmos to efficiently store and search multiple embeddings, including CLIP and color vectors, in a single collection. This provided the foundation for their exact color search and other retrieval needs. The results included sub-second search latency, a 79% reduction in element processing time, and predictable operations, enabling Cosmos to deliver a fast and precise search experience for its users.

Cosmos