Afya-Predict
A modular AI platform for predicting disease outbreaks across Tanzania.
Afya-Predict (Afya is Swahili for "health") is a modular, extensible platform for predicting disease outbreaks in Tanzania. It ingests health-facility, climate, and mobility data and surfaces early-warning signals for public-health decision-makers. The architecture is built around plug-in data sources so the same system scales across regions and disease categories instead of being rebuilt for each one.
afya-predict forecast --region dar-es-salaam --horizon 14d
sources ▸ facility · climate · mobility
features ▸ 42 engineered
signal ▸ elevated risk · week 3
output ▸ early-warning brief
// engineering notes
Decisions behind it
01
Plug-in data sources
Health-facility reports, climate series, and mobility signals each enter through the same connector interface. A new region or a new feed is a plug-in, so coverage grows without touching the modelling core.
02
Built for decision-makers, not dashboards
The output that matters is an early-warning signal a public-health team can act on — which means calibration and lead time are the metrics under scrutiny, not raw accuracy.
03
Scales across disease categories
Disease-specific logic is isolated from ingestion and feature engineering, so extending the platform to another category reuses the entire pipeline underneath it.