NightWatch — AI-Powered Real-Time Event Monitoring
Built an end-to-end AI event pipeline: RTSP ingestion -> frame sampling -> Gemini Vision reasoning -> structured event generation -> real-time alerting -> persistence. Integrated multimodal LLM reasoning with backend services and external notification APIs to automatically detect intrusions, loitering, and restricted-zone events without continuous human monitoring. Designed the application around asynchronous processing, failure isolation, event persistence, and real-time delivery so AI inference, alerts, and historical event workflows could evolve independently.
// INTERVIEW ENGINE