Open geospatial AI infrastructure

Earth observation data that works
for the contexts that matter most

OpenEarthStack builds open data infrastructure, geospatial AI tools, and research for anyone working in African and underrepresented contexts. We start where the gaps are largest.

What we do Get in touch

What we do

Geospatial AI research

We study where geospatial AI models break down in African contexts and why. Most models were built on data from Europe and North America. We work on closing that gap through rigorous evaluation and open research.

Open data for Africa

We curate and publish Earth observation datasets for African cities and landscapes in formats that actually work on the ground. Accessible without large downloads, documented in plain language, and free to use.

Technical consulting

We work with organisations that need geospatial AI applied to African contexts. Whether that is disaster response, agriculture, urban mapping, or climate monitoring, we help teams get from raw satellite data to something useful.

Tools and workflows

We build simple tools and write up the workflows that help practitioners use geospatial AI without needing a machine learning background. If you have satellite data and a problem to solve, we want to make that easier.


Why this matters

The infrastructure exists. It just wasn't built for here.

Satellite data now covers every part of the planet. AI can interpret it at scale. The tools exist and many of them are free. But they were built using data from places where labeled datasets are abundant, internet is fast, and the problems have already been studied for decades.

African cities, landscapes, and communities are largely missing from the datasets these models learned from. When you deploy them in informal settlements, smallholder farm contexts, or during a wet season with heavy cloud cover, they quietly perform worse. Not always visibly. But enough to matter when the output is informing a real decision.

For teams working on disaster response, food security, urban planning, or climate work in Africa, this is not an abstract research gap. It shows up in the work every day.

~3%
Of Earth observation deep learning datasets explicitly cover African contexts, from a review of over 400 datasets
Low
Representation of African cities in the urban flood, land cover, and building damage benchmarks used to evaluate AI models
4M+
People displaced by floods across African countries in 2024, with response systems that rely on data tools not built for those contexts

Working in Africa?
Let's build the infrastructure it needs.

Get in touch →