IPBMS — AI Patient Monitoring (Whammy Tech)
AI real-time patient-monitoring system detecting falls and seizures from live camera streams.
Overview
IPBMS is a real-time patient-monitoring system that watches live camera streams and automatically detects falls and seizures — an extra set of eyes for wards where continuous human observation isn't feasible. As AI/Backend Engineer, I developed the RTSP ingestion pipeline in OpenCV, YOLO pose-based fall detection, and a skeleton-motion module for seizure detection. To keep the system trustworthy in a clinical setting, I built an event-based alert architecture (Normal / Warning / Danger) with 5-frame validation that cuts false positives, JSON detection logs for auditability, and real-time notifications over WebSocket and Supabase Realtime. Patients, alerts and behavior logs are persisted in a scalable PostgreSQL schema. The emphasis throughout was reliability under real conditions: reducing false alarms without missing genuine emergencies, and delivering alerts fast enough for staff to act.
Tech stack
Have a similar project?
Tell us about it — from idea to production, the Zillexa team ships with you.