A nutrition assistant designed for Africa: the user takes a photo of their meal and instantly gets the dish identified (thiéboudiène, mafé, kedjenou, alloco…), an estimate of its nutritional value (calories, protein, carbs, fat), and personalized recommendations through a conversational agent — text or voice.
Built at Data Afrique Hub, where I currently work, on the AI-for-health side.
The project is split across three complementary repos:
- afrinutri — the app (Flutter + FastAPI backend)
- nutri-ia-data-collection — data collection, cleaning, and preparation (meal photos, nutritional data)
- nutri-ia-model-training — model training and fine-tuning (vision and embeddings)
Stack
- Flutter — mobile app
- FastAPI — backend
- Computer vision — identifying the dish from a photo
- Conversational agent (text + voice) — personalized recommendations