Every patient deserves follow-up.
RAYA is an AI platform that makes automated follow-up phone calls to patients in Darija (Moroccan Arabic). Doctors configure questions per patient, set call frequency, and the AI handles the rest — speaking naturally in Moroccan Arabic. After each call, the AI generates a risk evaluation, but the doctor always makes the final decision.
Features · Screenshots · How It Works · Tech Stack · Quick Start · Project Structure
In Morocco, 700 doctors emigrate every year. With only 6.7 doctors per 10,000 people, patients discharged after surgery, chemotherapy, or a diabetes crisis go home with zero follow-up calls. Complications get missed. They end up back in the ER — or worse.
RAYA solves this by automating follow-up calls in the language patients actually speak: Darija.
- AI Voice Calls in Darija — RAYA calls patients and speaks naturally in Moroccan Arabic using Edge TTS (
ar-MA-MounaNeural) - Doctor-Configured Questions — Doctors choose exactly what to ask each patient, with condition-specific presets for 7+ conditions
- Any Medical Condition — Diabetes, Alzheimer's, Chemotherapy, Post-Surgery, Heart Failure, Asthma, COPD, and more
- AI Risk Analysis — After each call, MiniMax LLM analyzes the conversation and generates a risk score with per-question concern levels
- Doctor Decision — The AI suggests, but the doctor decides: Approve, Escalate (create alert), or Dismiss
- Smart Call Queue — Automated daily scheduling based on call frequency (daily, weekly, every 3 days)
- Real-time Translation — Darija → English and Darija → French live translation
- Alert System — Severity-based alerts when doctors escalate cases
- Analytics Dashboard — Patient recovery rates, call volumes, risk trends
All patients with conditions, call schedules, status, and city — filterable by status.
Doctor configures patient info, condition (with presets), call schedule, and AI questions in a guided 3-step flow.
Today's calls at a glance — who's overdue, who's ready, who's completed. One-click to start a call.
RAYA speaks Darija in real-time. The right panel shows the doctor's configured questions being asked one by one. After the call, AI generates a full risk evaluation.
┌────────────────────────┐
│ 1. Doctor Configures │ Choose questions, call time, frequency per patient
└──────────┬─────────────┘
▼
┌─────────────────────┐
│ 2. RAYA Calls │ AI phones the patient, speaks Darija naturally
│ in Darija │ Asks each question, handles confused/unclear answers
└──────────┬──────────┘
▼
┌─────────────────────┐
│ 3. AI Analyzes │ Risk score + per-question concern levels
│ the Call │ Translates transcript to English/French
└──────────┬──────────┘
▼
┌─────────────────────┐
│ 4. Doctor Decides │ Approve (no action) / Escalate (create alert) / Dismiss
└─────────────────────┘
| Layer | Technology | Purpose |
|---|---|---|
| Frontend | Next.js 16, TypeScript, Tailwind CSS v4 | Modern responsive dashboard |
| Backend | FastAPI, SQLAlchemy, SQLite | REST API, ORM, lightweight DB |
| AI / LLM | MiniMax (primary), Groq Llama 3.3 70B (fallback) | Darija conversations, translation, risk analysis |
| Voice TTS | Edge TTS ar-MA-MounaNeural |
Moroccan Arabic text-to-speech (free) |
| Voice Fallback | ElevenLabs | Multilingual Arabic voice synthesis |
| Dev Tool | Cursor AI IDE | Full-stack app built with AI in 48 hours |
- Python 3.10+
- Node.js 20+
git clone git@github.com:itselcid/RAYA-AI.git
cd rayacd backend
# Install dependencies
pip install -r requirements.txt
# Configure environment — copy the example and add your API keys
cp .env.example .env
# Edit .env with your keys: MINIMAX_API_KEY (required), GROQ_API_KEY (recommended)
# Seed demo data (creates demo doctor account + sample patients)
python3 seed_demo.py
# Start the server
python3 -m uvicorn app.main:app --host 0.0.0.0 --port 8000 --reloadcd frontend
# Install dependencies
npm install
# Start dev server
npm run devGo to http://localhost:3000
Login with the demo account:
- Email:
doctor@test.com - Password:
test1234
| Service | Purpose | Get it at | Required? |
|---|---|---|---|
| MiniMax | LLM (Darija conversations + risk analysis) | https://platform.minimaxi.com | Yes |
| Groq | Fast LLM fallback | https://console.groq.com | Recommended |
| ElevenLabs | TTS voice fallback | https://elevenlabs.io | Optional |
Note: Edge TTS (Moroccan Arabic voice) works without any API key — it's free and built-in.
raya/
├── backend/ # FastAPI Python backend
│ ├── app/
│ │ ├── api/ # REST endpoints (patients, calls, simulate, alerts)
│ │ ├── models/ # SQLAlchemy ORM models (Patient, Call, CallMessage, Alert)
│ │ ├── schemas/ # Pydantic request/response schemas
│ │ ├── services/ # Business logic
│ │ │ ├── minimax_service.py # LLM chat + translation + risk analysis
│ │ │ ├── edge_tts_service.py # Moroccan Arabic TTS (primary)
│ │ │ ├── elevenlabs_service.py # ElevenLabs TTS (fallback)
│ │ │ └── groq_tts_service.py # Groq TTS (experimental)
│ │ └── prompts/ # AI system prompts for Darija medical agent
│ ├── seed_demo.py # Demo data seeder
│ └── requirements.txt
├── frontend/ # Next.js 16 TypeScript frontend
│ └── src/
│ ├── app/
│ │ └── (dashboard)/ # Dashboard pages
│ │ ├── dashboard/ # Overview with stats
│ │ ├── patients/ # Patient management
│ │ ├── queue/ # Daily call queue
│ │ ├── calls/ # AI call history + analysis
│ │ ├── simulate/ # Live AI call simulator
│ │ ├── alerts/ # Alert management
│ │ └── analytics/ # Charts and trends
│ ├── components/ # Reusable UI components (sidebar, logo)
│ └── lib/ # API client, utilities
├── screenshot_*.png # App screenshots for presentation
├── qa_cheatsheet.md # Q&A guide
- B2B SaaS model — sell to hospitals and clinics per patient/month
- Market: 32M Moroccans on AMO insurance, 150+ hospitals, 1000+ clinics
- ROI: 1 prevented readmission = 15,000 – 50,000 MAD saved
- Expansion: Morocco → Gulf → France (diaspora) → Sub-Saharan Africa
Built at CURECODE Hackathon UM6P - Faculty of Medical Sciences & UM6P Hospitals x 1337 Benguerir 2026