AI-Driven Guest Identification and Photo Retrieval System

There is a specific kind of let-down that follows big events. You know photos were taken. You were there. A professional was working all day. Two weeks later a gallery link arrives, 3,000 images in it, and the expectation is that each of the two hundred guests will scroll through until they find themselves. Most give up within ten minutes. The photos that were taken of them — professionally, at something they cared about — they never actually see. We built a system so that doesn't have to happen anymore. Guest scans a QR code at the venue entrance, uploads a selfie, and a personal email arrives after the event with every photo from the day where their face appeared. DeepFace handles recognition, MTCNN handles detection, Node.js runs the backend, cloud storage absorbs the volume. We tested it at four events: two weddings, a university tech festival, a corporate town hall. Matched at 95 percent accuracy, processed around 25 photos per second on a standard laptop. This paper describes how it works, what failed during development, what we changed, and where the gaps remain.