Inspiration

During midnight pediatric emergencies and rural network blackouts, single parents and frontline responders face life-threatening delays when cloud-only health apps fail without internet. We were inspired to build CareSync AI so life-saving medical triage, drug-allergy protection, and emergency dispatch work seamlessly both WITH Net and WITHOUT Net in English and Tamil.

What it does

CareSync AI is an offline-first, 10-module enterprise clinical triage and emergency command ecosystem built for all age groups (Child 0–12y, Adult 13–59y, and Senior 60+):

  • Master Unified Command Center (/master-command.html): Real-time 10-service health telemetry radar with hands-free bilingual voice SOS (Tamil & English).
  • Universal Triage & Drug Safety Shield (/index.html): Hybrid AI triage, Holliday-Segar hydration (ml/day), weight-based pediatric/senior dosage verification, and automated drug-allergy blocking with IndexedDB offline auto-sync.
  • Family Health Vault & Google FHIR Hub (/family-hub.html): Permanent blood group/chronic profile vault, personalized care plans, and Google Cloud Healthcare FHIR-Lite JSON export.
  • 108 Ambulance & Hospital ICU Dispatch (/emergency-command.html): Live hospital ICU bed availability and priority pre-arrival dispatch tickets (EMS-2026-XXXX).
  • Autonomous Agentic AI Autopilot (/agentic-command.html): 5-step autonomous clinical agent that analyzes fever trends, screens drug allergies, reserves ICU beds, and dual-syncs records to Firebase Cloud Firestore and Spring Boot SQL.
  • Offline SOS QR & Vitals Telemetry (/vitals-analytics.html): Pure-Canvas zero-internet scannable QR Medical Passport for paramedics and live temperature progression charts (~38ms On-Device vs ~620ms Cloud).
  • Multimodal Vision Medicine & Wound Scanner (/vision-scanner.html): Medicine bottle label and skin/burn photo triage paired with Java pharmacovigilance safety audits.
  • Grounded WHO Clinical RAG Lab (/genai-rag-lab.html): Zero-hallucination WHO/Red Cross protocol retrieval grounding with Socratic multi-turn memory and structured JSON tool outputs.
  • Live Community Outbreak Radar (/outbreak-radar.html): Real-time regional symptom cluster surveillance and animated bio-surveillance heatmap.
  • Pure-Java Predictive ML Lab (/ai-ml-predictor.html): Logistic Regression Sigmoid risk probability (%), OLS Linear Regression 2-hour future fever forecasting (°F), and Cosine Similarity clinical vector matching.

How we built it

  • Frontend & Offline Engine (28 Modular Files): Engineered 10 HTML5 Enterprise Portals, 9 responsive dark-mode CSS3 design systems, 11 ES6+ JavaScript engines, an IndexedDB Offline Clinical Vault, a PWA Service Worker (sw.js), HTML5 2D Canvas QR & Radar generators, and Web Speech Bilingual Voice APIs.
  • Hybrid AI & Cloud Layer: Integrated Firebase AI Logic SDK (InferenceMode.PREFER_ON_DEVICE), Cloud Gemini 2.5 Flash, On-Device Gemini Nano, Firebase Cloud Firestore, and Google Healthcare FHIR-Lite JSON.
  • Enterprise Backend (34 Java Classes): Built with Java 17, Spring Boot 3 (10 REST Controllers), Spring Data JPA, and H2 Relational SQL.

Challenges we ran into

  • Designing deterministic offline clinical fallbacks that match cloud accuracy during complete internet blackouts without hallucinating medical dosages.
  • Synchronizing IndexedDB offline queues cleanly across both Spring Boot SQL and Firebase Cloud Firestore without duplicating patient records upon reconnection.

Accomplishments that we're proud of

  • Architecting a complete 10-controller Java enterprise backend paired with 28 modular frontend files and sub-40ms offline triage resilience.
  • Implementing pure-Java predictive ML algorithms (Logistic Sigmoid, OLS Linear Regression, Cosine Similarity) and zero-hallucination WHO Clinical RAG grounding from scratch.

What we learned

  • How InferenceMode.PREFER_ON_DEVICE unlocks true low-latency resilience (~38ms vs ~620ms cloud) for mission-critical healthcare tools in low-connectivity regions.
  • How combining deterministic clinical guardrails (drug allergies and weight limits) with Generative AI creates trustworthy, safe medical decision support.

What's next for CareSync AI: Offline-First Clinical Triage Ecosystem

  • Integrating live hospital HL7/FHIR endpoints for direct emergency room pre-registration.
  • Adding SMS emergency fallback gateways for rural feature-phone users and expanding support across additional Indian regional languages.

Built With

Share this project:

Updates

Submission history