Conntour raises $7M from General Catalyst, YC to build an AI search engine for security video systems

Conntour's funding signals a critical shift: unstructured video data is becoming a first-class citizen in enterprise data stacks. As data engineers, we've long treated video as a storage problem—expensive, difficult to index, mostly archived. This product suggests the economics have flipped. Natural language interfaces over video feeds mean we'll need to architect pipelines that feed embeddings into semantic search systems rather than traditional schemas. The implication is architectural: teams will need to decide whether to build in-house video processing with models like CLIP or YOLO, or integrate APIs like Conntour. This mirrors the embedding-as-infrastructure trend we've seen with text search. What matters most is that security teams now expect ad-hoc querying of historical video—the same way analysts query data lakes. That creates new demand for streaming infrastructure, vector databases, and real-time feature pipelines. My recommendation: start cataloging video metadata through your analytics layer now, even if you're not deploying semantic search yet. Understanding frame rates, resolution, and storage costs will inform whether building or buying makes sense for your infrastructure.