Local-language audio, images and labels Rights-cleared audio, images and on-the-ground labels in Nigerian languages, collected by people who actually speak them — each datapoint tagged with who, when, where, and consent.
100%of every batch is escrow-backed before a single datapoint is collected
₦0held by hannu — licensed partners hold the money, we keep the record
1 zonelive today — Lagos, Nigeria, expanding as the gates hold
You train on data you cannot trace — and cannot defend when it counts.
Scraped and synthetic data comes with no record of who collected it, whether anyone consented, or whether the rights were ever cleared — and it drifts further from reality with every model that leans on it. When a model audit or a licensing negotiation asks where a label came from, "a crowd platform, somewhere" is not an answer.
The same anonymous account can label your set twice, and a Nigerian language can come back wrong from someone who never spoke it. What closes the gap is real, rights-cleared ground truth collected by trained, verified people — every datapoint carrying who collected it, when, where, and the consent behind it.
One lifecycle, funded to signed. Fund it once — the platform runs every step and hands back one signed report.
₦ held
Define & fund
Set the data you need — the languages, the labels, the proof each datapoint must carry — and fund it once. The money locks in escrow before any collection starts.
Wallet & escrowThe checks you can run. Each one honestly stated — Live, Pilot or Gated, never claiming more than it delivers.
Human labels and evaluation sets Training and evaluation labels scored inside every batch by seeded known-answer checks, so quality is measured on the way in — not discovered after you have trained on it.
Curated datasets, provenance intact A packaged, rights-cleared dataset delivered as one signed record with full consent lineage — running in pilot with early data partners.
Go deeper
Guide
What is inside a signed record
The checks we run, the sources matched, and the signed provenance your own systems can re-verify.
GuideA trained, certified workforce
Operators are trained and certified before they collect, so local-language labels come back right the first time.
GuideA proven collector track record
Every Operator carries a portable record of the work they have passed — so you know who stands behind each datapoint.
Put ground truth under your models.
Real, rights-cleared African data — collected by trained, verified people and returned as one signed record you can trace.