Frostbyte Hackathon 2026 · Healthcare & BioTech Theme · Explainable AI

A full-stack web application that helps patients record daily health data and caregivers monitor and prioritize patients using SHAP-powered Explainable AI risk assessment.

Note: This is a decision support system. It does not perform medical assessments or provide care recommendations.

Healthcare caregivers managing multiple patients need an efficient way to monitor health data and identify patients who may require closer attention. Patients need a simple way to record daily vitals and understand their health status.

SmartCare provides a complete ecosystem for remote patient monitoring:

• For Patients: Easy daily health data recording with immediate risk level feedback, SHAP-based Explainable AI insights explaining why they are at risk, and a Health Score (0-100).
• For Caregivers: A Command Center dashboard featuring analytics summary cards, risk distribution charts, critical alerts, and patient management tools.
• Real-World Utility: Caregivers can generate and download PDF Medical Reports to share with doctors.
• Bilingual & Accessible: Beautifully animated UI (Framer Motion) fully localized in English and Thai (��/��).
• Bilingual & Accessible: Beautifully animated UI (Framer Motion) fully localized in English and Thai (/).
• Real-time Monitoring: Automatic polling with toast notifications when new health data arrives.

┌──────────────────────────────────────────┐
│           Next.js Application            │
│  ┌──────────┐   ┌─────────────────────┐  │
│  │    UI    │   │  API Route Handlers │  │
│  │ (React)  │   │  /api/auth          │  │
│  │          │──▶│  /api/records       │  │
│  │          │   │  /api/patients      │  │
│  │          │   │  /api/predict ──────│──┼──▶  Python FastAPI
│  └──────────┘   └─────────┬───────────┘  │      AI Service
│                           │              │      POST /predict
│                           ▼              │
│                      PostgreSQL          │
│                     (Prisma ORM)         │
└──────────────────────────────────────────┘

We compared 3 algorithms and selected Random Forest for its excellent balance of accuracy and SHAP explainability:

  temperature: 0.3935 ███████████████████
     systolic: 0.2155 ██████████
   heart_rate: 0.1913 █████████
    diastolic: 0.1294 ██████
      symptom: 0.0704 ███

Every prediction comes with:

• SHAP contribution values for each health feature
• Visual bar chart showing which factors increase or decrease risk
• Human-readable explanation in both English and Thai
• Health Score (0-100) based on risk level and model confidence

• ✅ Daily health data recording (temperature, heart rate, BP, symptoms)
• ✅ AI risk assessment with SHAP explanation bar chart
• ✅ Health Score gauge (0-100)
• ✅ Health trend charts (last 14 records)
• ✅ Paginated health record history

• ✅ Analytics Dashboard with summary cards (Total / High / Medium / Low)
• ✅ Risk Distribution pie chart
• ✅ Critical Alerts section with pulsing indicators for high-risk patients
• ✅ Patient management (add, edit, delete)
• ✅ Caregiver observation notes
• ✅ PDF Medical Report export
• ✅ Real-time polling (30s) with toast notifications

• ✅ Bilingual UI (English / Thai) with AI explanations in both languages
• ✅ JWT authentication with role-based access
• ✅ Responsive design (mobile & desktop)
• ✅ Framer Motion animations throughout

User ──── Patient ──── HealthRecord ──── Prediction
             │
             └── Note (Caregiver observations)

   HealthRecord Fields:
   - temperature, heartRate, systolic, diastolic, symptom

Models: User, Patient, HealthRecord, Prediction, Note Roles: PATIENT, CAREGIVER

Python AI Service:

• Node.js 18+
• PostgreSQL
• Python 3.9+

cd smartcare
npm install

# Update DATABASE_URL in .env with your PostgreSQL credentials
# Example: postgresql://user:password@localhost:5432/smartcare

createdb smartcare
npx prisma migrate dev --name init
npm run seed

npm run dev
# Opens at http://localhost:3000

cd ai-service

# Create virtual environment
python3 -m venv venv

# Activate virtual environment  
source venv/bin/activate

# Install dependencies
pip install -r requirements.txt

# Train the model (generates SHAP explainer + metrics)
python train_model.py

# Start the service
uvicorn main:app --reload --port 8000

1. Landing Page → Modern landing with stats, how-it-works, and tech stack
2. Login → Sign in as patient1@demo.com
3. Dashboard → See current health status, health score, and risk level
4. Record Health Data → Fill vitals, submit, see AI assessment
5. SHAP Analysis → View contributing factors bar chart with explanations
6. Switch to Caregiver → Login as caregiver@demo.com
7. Command Center → Analytics cards, pie chart, critical alerts
8. Patient Detail → Full history, charts, notes, PDF export
9. Switch Language → Toggle EN/TH in navbar

smartcare/
├── src/
│   ├── app/
│   │   ├── login/              # Login page
│   │   ├── patient/
│   │   │   ├── dashboard/      # Patient dashboard + health score
│   │   │   └── new-record/     # Health data form + SHAP chart
│   │   ├── caregiver/
│   │   │   ├── dashboard/      # Analytics command center
│   │   │   └── patients/[id]/  # Patient detail + PDF export
│   │   └── api/
│   │       ├── auth/           # Login/logout
│   │       ├── records/        # Health records
│   │       ├── patients/       # Patient data
│   │       └── predict/        # AI prediction proxy
│   ├── components/
│   │   ├── Navbar.tsx          # Navigation + language toggle
│   │   ├── RiskBadge.tsx       # Risk level indicator
│   │   ├── HealthForm.tsx      # Health data form + SHAP display
│   │   ├── ChartPanel.tsx      # Recharts health trends
│   │   ├── LandingClient.tsx   # Landing page
│   │   └── LanguageProvider.tsx # i18n context
│   ├── lib/
│   │   ├── auth.ts             # JWT utilities
│   │   ├── db.ts               # Prisma client
│   │   └── translations.ts     # EN/TH translations
│   └── middleware.ts           # Route protection
├── prisma/
│   ├── schema.prisma
│   └── seed.ts
├── ai-service/
│   ├── main.py                 # FastAPI app + SHAP
│   ├── train_model.py          # Model training + comparison
│   ├── requirements.txt
│   ├── model.pkl               # Trained model
│   ├── shap_explainer.pkl      # SHAP TreeExplainer
│   └── model_metrics.json      # Training metrics
└── README.md

• Project Title: SmartCare
• Theme: Healthcare & BioTech
• Team Member:
  1. Tanakorn Kaewmai - Solo Developer / Full Stack & AI
• Originality Statement: All work in this repository is original and created specifically for the Frostbyte Hackathon.
• Datasets Used: Synthetically generated clinical data using train_model.py based on domain-informed distributions for vital signs and symptom patterns.
• External Libraries & Citations:
  • Frontend: Next.js, React 19, Tailwind CSS, Framer Motion (UI animations)
  • Backend/Database: Prisma ORM, PostgreSQL, JSON Web Tokens (jose)
  • AI & Data Science: Python, FastAPI, scikit-learn (Random Forest), XGBoost, SHAP
  • PDF Export: react-to-print

• Real-time WebSocket notifications (upgrade from polling)
• Wearable device integration (Apple Watch, Fitbit)
• Time-series forecasting for health trends
• Multi-hospital deployment with admin panel
• Doctor role with treatment recommendations
• HIPAA compliance features