PulseAI

PulseX: AI-powered health companion analyzing your Apple Watch data to deliver personalized insights, proactive alerts, and emotional support—right on your wrist.

Comment

Inspiration

Healthcare today is built on snapshots—expensive, time-consuming office visits that miss 98% of your daily health story. We wanted to bring continuous, objective data into the doctor’s toolkit and empower patients to understand their own bodies in real time.

What it does

  • Continuous Monitoring: Ingests heart rate, sleep, activity and stress data from your Apple Watch every minute.
  • Smart Dashboard: Visualizes daily/weekly trends, deep-sleep stages, mood logs.
  • Proactive Alerts: Scans hourly for anomalies; critical events trigger instant emails to your doctor.
  • One-Click Reports: Bundles your key metrics into a PDF summary you can share with your care team.
  • AI Doctor & Therapist Chatbot: Cross-references your vitals and symptoms to deliver personalized advice—or simply talk you through anxiety like a friend.

How I built it

  1. Data Collection & Ingestion: Health Auto Export pushes JSON payloads every minute to a Node.js server.js endpoint.
  2. Storage: Parsed metrics are stored in PostgreSQL tables (realtime vitals, sleep sessions, mood logs).
  3. Backend API: Express.js endpoints run SQL queries and expose JSON to the frontend.
  4. Frontend Dashboard: React app fetches API data to render live charts and trend views.
  5. Scheduling & Automation:
    • Hourly Cron Jobs: Anomaly detection + email alerts.
    • Monthly Jobs: PDF report generation + automated email.
  6. AI Insights: Select metric snapshots sent to ChatGPT API with custom system prompts—responses power the Doctor/Therapist chatbot.

Challenges I ran into

  • Real-Time Sync: Ensuring Health Auto Export → server → database with sub-minute latency.
  • Anomaly Detection Algorithms: Tuning false-positive rates on noisy biometric data.
  • Privacy & Security: Encrypting data in transit and at rest while using third-party APIs.
  • Prompt Engineering: Crafting system prompts that yield accurate, user-specific health advice.

Accomplishments that I am proud of

  • Built a fully automated end-to-end pipeline (watch → DB → dashboard → alerts → AI).
  • Achieved <60 s data round-trip latency for real-time insights.
  • Launched our AI Doctor/Therapist chatbot with personalized recommendations.
  • Automated PDF reporting and doctor-notification workflows.

What I learned

  • The power—and pitfalls—of real-time health data: balancing granularity with noise reduction.
  • How to design clean, intuitive dashboards that distill complex metrics.
  • Best practices for secure data handling in a health context.
  • The art of prompt engineering for domain-specific AI advice.

What’s next for PulseAI

  • Predictive Population Analytics: Leverage anonymized data from millions of users to flag early disease patterns (e.g., cancer risk) and send preemptive “red-flag” alerts.
  • Personalized Lifestyle Coach: Auto-generate your ideal diet, sleep, and work-rest routines based on your unique biometrics—and refine them with continuous feedback.
  • Native Mobile App: Roll out an iOS/Android version for richer push notifications, voice check-ins, and one-tap chatbot access.
  • Research & Partnerships: Offer a secure, de-identified health data platform for clinical studies and AI model improvements in collaboration with healthcare institutions.

Built With

Share this project:

Updates