Electric Sheep Africa

Africa's ML dataset infrastructure.

We build and maintain the largest open collection of machine-learning-ready datasets on the African continent — 7,900+ datasets spanning health, agriculture, energy, finance, education, governance, and more, covering all 54 countries.

What makes ESA datasets different

Raw data for Africa exists. What hasn't existed — until now — is a single, consistent, ML-ready layer on top of it.

Every dataset in this collection goes through ESA's data pipeline:

Cleaning Missing value markers unified across formats (N/A, null, none, -, unknown, no data, #N/A → NaN). Columns with >80% missingness removed. Duplicate rows and malformed entries resolved.

Normalisation Column names lowercased and snake_cased. Datatypes enforced. Units and categorical encodings standardised across datasets from the same source family.

Augmentation Datasets enriched with contextual features — geographic identifiers, temporal markers, cross-source linkage keys — where source data is sparse or inconsistently structured.

Provenance tracking Every row carries esa_source and esa_processed fields. Every dataset ships with a full BibTeX citation traceable to the original publisher.

ML formatting 80/20 train/test splits using a fixed random seed (42). Saved as Snappy-compressed Parquet. Loadable in one line via the datasets library.

We ingest from the Humanitarian Data Exchange (HDX), Our World in Data (OWID), CGAP, and other primary sources — and do the work that turns raw development data into something a researcher or engineer can actually use.

What we cover

Research & models

Beyond datasets, we build:

• Chewie / Humani — a MedGemma-based clinical decision support model for community health workers
• Genomics research — SSA-specific breast cancer and gene expression datasets (genomics.electricsheep.africa)
• Sector simulations — the Nigerian Economic Policy Simulator and other applied AI tools

Use our data

from datasets import load_dataset

ds = load_dataset("electricsheepafrica/DATASET_NAME")
train = ds["train"].to_pandas()

To request a dataset not yet in the collection, email kossi@electricsheep.africa.

Our 3-year research roadmap

Year 1 — Foundation & contextual intelligence Building the dataset infrastructure, contextualised models, and research pipelines that establish ESA as Africa's ML data layer.

Year 2 — Systems, deployment & simulation Moving from data to deployed systems: clinical AI, economic simulators, sector-specific models grounded in ESA datasets.

Year 3 — Scale & spin-out Spinning out products and policies built on the foundation. Expanding beyond Nigeria to a continental mandate.

Support our work

ESA is a non-profit. If our datasets have been useful to your research or product, consider supporting us.

Nigeria — Kuda Bank · Account: 3003437130 · Electric Sheep Africa United States — Lead Bank · Account: 217143145453 · ACH/Wire Routing: 101019644 · Bank address: 1801 Main St., Kansas City, MO 64108

Electric Sheep Africa · Lagos, Nigeria · electricsheep.africa