UUAMNI builds the data layer for AI in the languages it has barely learned. We start with Igbo and extend to other underrepresented African languages.
AI does not finish itself. It needs people to keep correcting it, ranking its outputs, and teaching it what good looks like. That work does not stop, and taking the people out of it only makes the system worse. We build where the teachers are.
Uche, our annotation platform, is staffed by language educators: people who can both speak a language and explain it. They produce translation, preference, and evaluation data with a level of fluency that scraped text cannot match.
A gated Igbo translation sample is in preparation as a public reference for the quality and methodology behind this work. Follow the organization to see it when it lands.