About the Project

Problem Statement & Motivation

Access to accurate and timely health information is a major global challenge, particularly in developing regions such as sub-Saharan Africa. Millions of people face barriers when trying to understand medical information due to limited digital resources, lack of infrastructure, and especially language differences.

In many communities, healthcare guidance is often available only in dominant global languages such as English, which excludes large populations who primarily communicate in local languages like Amharic, Tigrinya, and Oromo. This language gap becomes even more critical in sensitive health domains such as maternal health, reproductive health, infectious diseases, and mental health, where misunderstanding can lead to serious consequences.

Existing AI language models are primarily trained on high-resource languages and struggle with low-resource African languages. As a result, they often fail to generate accurate, fluent, and culturally appropriate responses for users in these regions.

This project addresses these limitations by developing a multilingual healthcare question-answering system using Retrieval-Augmented Generation (RAG). The system retrieves relevant medical knowledge from curated datasets and generates context-aware, language-consistent responses.

The goal is to build a system that can understand a user’s question in English or Amharic and respond accurately in the same language, ensuring accessibility, clarity, and trust in AI-driven healthcare support tools.

Ultimately, this work aims to support real-world applications such as digital health assistants, patient education tools, and clinical decision support systems for healthcare workers in rural and underserved communities.