VitaSensor

VitaSensor is a full-stack smart-hospital monitoring system that tracks patients' vital signs in real time (heart rate, SpO2, temperature, ECG) and uses machine learning to flag deterioration before it becomes an emergency.

Context and problem. In a hospital, the window between a patient's condition worsening and a clinician noticing can be critical. Raw monitors show what's happening now; they don't anticipate what's coming, and they don't route the right information to the right person. The challenge here was twofold: stream live data reliably to the people who need it, and turn that stream into something predictive rather than merely reactive.

Solution. The platform is built in three layers. A React + TypeScript frontend gives each role — admin, doctor, nurse — its own dashboard, updated live through Socket.IO. A Node.js/Express backend handles authentication (JWT), REST routes, and the real-time event layer, persisting everything to PostgreSQL through Prisma. A dedicated Python AI service runs the models: an LSTM that forecasts SpO2 thirty minutes ahead, a Random Forest that classifies patient status, an Isolation Forest for anomaly detection, and logistic regression for cardiac-risk scoring. On top of this sits a multi-level alert engine (critical/high/medium/low) with an acknowledgment-and-intervention workflow so alerts are actioned, not just displayed.

Engineering and quality focus. The system is intentionally organized around clear boundaries — the AI service is decoupled from the API, and the real-time layer is separated from the request/response layer — which keeps each part testable and independently evolvable. The README is honest about scope: some pieces (full Python-service integration, SMS/email notifications, multi-hospital support) are on the roadmap rather than shipped, documented openly rather than implied.