ShiftSight

ShiftSight is an AI closing-shift verifier for restaurants and hotels. Closing procedures in hospitality depend on memory and paper checklists at the end of a long shift. I’ve worked restaurant closes and seen how one missed item, like a bottle left on the bar, a cooler door ajar or garnish left out, turns into a theft, safety or food-safety problem the next morning. A supervisor uploads a 30-60 second walkthrough video and picks a five-item closing checklist. ShiftSight marks each task Completed, Needs review or Not visible, with an annotated evidence frame and timestamp, and flags issues the checklist didn’t cover. The manager confirms or rejects every finding, then gets a ready-to-send shift handoff or incident report. It’s a review assistant, not an autonomous inspector: nothing reaches the report without human confirmation. Built with React and FastAPI. Note: the demo runs in demo mode with simulated findings, because API billing wasn’t configured before the deadline. Video processing, evidence frames, the review flow, and the report work fully, and the repo includes the Meta model integration.