NLP-ICS/AfriMed_Tutor

A guideline-grounded study assistant for African medical students (AfriMed). AfriMed Tutor pairs an LLM with a small retrieval store of African clinical guidelines so that answers to medical questions are anchored in locally-relevant practice rather than Western-default training data. # AfriMed Tutor: Guideline-Grounded Medical Education Platform **A retrieval-augmented generation system delivering evidence-based clinical education to African medical students.** View Full Demo Video --- ## System Overview AfriMed Tutor addresses a critical gap in medical education infrastructure by providing **grounded, evidence-based learning support** tailored to the African clinical context. The system combines: - **Dense + Sparse Retrieval**: Hybrid retriever architecture over African medical guidelines - **Generative Question Answering**: LLM-synthesized responses with explicit guideline citations - **Interactive Quiz Engine**: MCQ generation and adaptive learning feedback - **Explanation Comparison**: Student reasoning validation against expert clinical protocols - **Automated Evaluation**: Groundedness and retrieval quality metrics for system monitoring **Technical Stack**: Python, FAISS, Streamlit, Anthropic/OpenAI APIs, Pydantic --- ## Key Features | Feature | Description | Impact | |---------|-------------|--------| | **Guideline-Grounded QA** | Answers clinical questions with explicit source citations from African medical guidelines | Ensures factual accuracy and trustworthiness for medical education | | **Retrieval Comparison** | Evaluates dense (FAISS) vs. sparse (BM25) retrieval strategies | Optimizes retrieval quality; identifies when supplementary retrieval is needed | | **Groundedness Evaluation** | Automated judge for answer-to-evidence alignment using LLM critique | Ensures generated responses remain faithful to source material | | **Multi-Modal Interface** | Streamlit UI supporting ask/quiz/explain modes plus CLI fallback | Accessible across devices and learning contexts | | **Extensible Architecture** | Modular design supports multiple LLM providers and embedding services | Production-ready deployment flexibility | --- ## Results & Impact - **High-Quality Baselines**: Comprehensive evaluation using AfriMed-QA dataset wi …