Case Study: Dragonfruit AI achieves real-time multi-camera computer vision at enterprise scale with Qdrant

Dragonfruit AI scales real-time multi-camera computer vision with Qdrant

Dragonfruit AI, a provider of enterprise-ready computer vision solutions, faced the challenge of scaling its real-time analytics platform across thousands of bandwidth-constrained retail and warehouse locations. Their needs included processing high-frame-rate video feeds on edge devices, tracking objects across multiple cameras with fast similarity search, and operating a vector database at a scale of hundreds of billions of vectors, all while meeting strict latency targets and keeping costs low. They chose the open-source Qdrant vector search engine to power this solution.

By implementing Qdrant, Dragonfruit AI leveraged its per-collection configurability to tune different workloads and used float16 vectors for improved memory efficiency. This provided the performance headroom for real-time retrieval and high-velocity writes, enabling use cases like multi-camera person re-identification and self-checkout verification. The results included a lower total cost of ownership, the ability to launch new AI agents faster, and the quantitative achievement of processing many cameras per site at high FPS with only compact inference data transmitted to the cloud.

Dragonfruit AI