Artificial intelligence
that speaks your languages.
Umbaji builds speech recognition, translation, and synthesis models for West African languages, starting from the communities who speak them and reaching the businesses who need them.
Spoken in
Working with
- Togo AI Lab
- AfricaNLP
Our capabilities
Language technology built for West Africa
Umbaji exists to close the linguistic digital divide. From raw voice to a benchmarked model, we cover the full chain: capture the language, structure it, and prove it works.
Explore our products
The Yodi platform
Language technology for every West African voice
Yodi is our voice AI model for speech recognition, translation, and synthesis, trained on data collected and validated by the communities who speak these languages.
- 50+Languages in progress
- 70+Contributors
- 45+Partnerships
- 4Datasets on Hugging Face
Our recent case studies
Each entry traces work carried out in the field, with the communities and institutions it was built with.
Umbaji is built with you
Whether you're a native speaker, a researcher, a company, or an institution, there's a place for you in what we're building.
Contribute your voice
Record, transcribe, or validate language data on the Data Hub. Accepted work gets paid.
Work with us
Internships, contracts, and open roles in robotics, cloud engineering, project management, and design.
Become a partner
Universities, governments, NGOs, and companies: let's build models useful to your field.
“We believe intelligence should speak the languages people actually live in.”
We picture a continent where speaking Ewe, Kabiyè, or any local language is never a barrier to reaching healthcare, education, or public services. Where AI models understand African languages as well as dominant ones, because Africans built them.
