HealthLit AI
An offline AI assistant that answers health questions in English and four Nigerian languages
Overview
Make trustworthy health information accessible in low-resource settings: no internet required, not limited to English, and answers grounded in recognised health sources.
The problem
Health information online is usually written in dense clinical English, assumes a constant connection, and is rarely available in Yoruba, Hausa, Igbo or Pidgin. In many Nigerian communities, that puts it out of reach.
What I built
An AI system that runs locally and searches a curated knowledge base of 12 WHO and Federal Ministry of Health topics before answering. Users can ask by text, voice or image, and get an answer in their chosen language. Patient data never leaves the device.
My role
Solo: built the retrieval pipeline over the health knowledge base, the Django backend, the React frontend, and connected speech, image and translation models around the core language model.
Technical challenges
Getting a small local model (3.8 billion parameters, no cloud fallback) to stay grounded in the source documents rather than making things up, and chaining four separate models (language, translation, speech, vision) into one offline pipeline without a GPU server.
Current status
Prototype. Built and presented as my final-year project (CSC 431, University of Ilorin, 2026). Not deployed as a public product.
What it does
- Accepts questions by text, voice or image
- Answers grounded in a 12-topic WHO/FMOH knowledge base covering malaria, diabetes, hypertension, maternal health and more
- Responds in English, Yoruba, Hausa, Igbo and Nigerian Pidgin
- Runs entirely offline on a standard laptop
- Patient data never leaves the device
Technology
Need something like this for your business?
Tell me what you're working on and I'll get back to you.