Cross-Camera Re-ID¶

Client: Maleakhi Wijaya, Tenyks.

Tenyks (YC S21) builds Vision AI agents for leading restaurant and hospitality brands. The ability to track anonymous customer journeys is mission-critical for consumer-facing businesses that rely on unreasonable hospitality and a consistent brand experience to attract and retain loyal customers. At Starbucks, for example, some customers waited 40 minutes for a latte, sales fell, and quarterly profit dropped almost 7x, from $909.3 million to $133.1 million. Today's computer vision systems can follow a person within a single camera's view, but keeping track of who is who as they walk from one camera to the next remains an unsolved problem. Your task is to build a multi-camera tracking and privacy-preserving person re-identification system that connects fragmented CCTV sightings into continuous, anonymous journeys. Joined-up journeys like these would let businesses see a customer's whole visit or a car's full path through a drive-through, spot where time is lost and queues build up, and plan staffing and training around what actually happens on the floor. The technical challenge is tracking people reliably through occlusion, very different camera angles, and views that sometimes overlap and sometimes leave gaps. Tenyks will provide real-world video footage, GPU compute, technical mentoring and access to AI development agents. All footage provided is used only with the explicit permission of Tenyks' clients for this purpose, in accordance with their consent and applicable data-use requirements.