selam-feyisa/seha-ethiopia

AI-powered healthcare document processing for Ethiopia # SEHA — AI Healthcare Assistant for Ethiopia An AI-powered healthcare assistant built for Ethiopia. ## Team - Fayza — AI & Backend Lead - Selam — Full-Stack & Data Lead ## Model Performance | Metric | Value | |--------|-------| | Model | Random Forest + XGBoost | | Overall Accuracy | 100% | | Test Set Size | 20% of dataset | | Fairness | No diseases below 5% accuracy threshold | | Top Symptoms | muscle_pain, yellowing_of_eyes, nausea, itching, dark_urine | ## Azure Services Used | Service | Purpose | |---------|---------| | Azure ML | Symptom model training and deployment | | Azure OpenAI (o4-mini) | Health Q&A via Ask SEHA | | Azure OpenAI (text-embedding-3-small) | RAG document embeddings | | Azure Document Intelligence | Medical document analysis | | Azure Blob Storage | MoH PDF document storage | | Azure AI Search | Vector search for RAG | ## 🚀 Local Setup Instructions ### Prerequisites - Python 3.10+ - Node.js 18+ - Azure account (university) with resources created (see Fayza) ### Backend Setup ```bash cd backend cp .env.example .env # Fill in your keys pip install -r requirements.txt uvicorn main:app --reload