EdgePilot AI

EdgePilot AI is an AI-powered safety monitoring platform designed to help operators identify and respond to meaningful safety events in industrial and warehouse environments. The working prototype uses browser webcam or uploaded video as its visual input and processes frames through Gemini Vision for object recognition, scene understanding, and safety-event detection. Detected events are passed through a deterministic risk engine that evaluates evidence, confidence, risk level, and event cooldowns before creating an incident. This prevents normal activity and repeated detections from unnecessarily generating incidents. For meaningful events, Groq provides contextual reasoning, a concise explanation, and a recommended action. The resulting incidents are stored in Supabase PostgreSQL and presented through a React and TypeScript dashboard, where operators can review incident details, risk levels, analytics, and system status. Supabase Authentication provides user login and protected application routes. The system follows a human-in-the-loop approach: AI assists with perception and reasoning while the human operator makes the final decision to acknowledge or escalate an incident. The prototype is deployed using Vercel for the frontend and Render for the backend. Future development could extend the platform toward physical edge-device deployment, additional sensors, multi-camera monitoring, and more advanced edge inference.