About
Namu AI Lab is the research unit of Namu, working on speech in Hausa and Zarma.
Namu builds speech-native models, datasets and products for African languages. The products live at namu-ai.org. The Lab publishes the research underneath them, as programmes, papers, datasets, evaluations and research demos. We start with Hausa, which has about 94 million first- and second-language speakers (Ethnologue, 2026), and Zarma, which has about six million (Ethnologue, 26th edition, 2023). We work in Niger first, because public Hausa speech data mostly comes from Nigeria and Zarma is almost absent from speech technology.
Our work runs in five programmes, from collecting consented speech to measuring what models do with it. Each programme page lists its next steps and ends with what we do not claim yet.
How we workFour commitments that come before any result
Consent before collection
No one records for the Lab without informed consent, given in their own language. Speakers choose which uses they allow, such as research, open release or voice synthesis, and they can withdraw. Their choices travel with every file made from their voice.
People are paid, and communities are partners
Everyone who records, transcribes or rates speech for the Lab will be paid for their time, at a rate we publish before collection starts. Speakers, reviewers and linguists will be credited as colleagues, and as co-authors where their work earns it.
A number comes with its conditions
Every result names its test set, its normaliser and version, and its listeners where there are any, and says what it cannot show. Hausa error rates come twice, strict and folded. We publish negative results too.
Open by default, closed for a reason
We release tools, test sets, pipelines and measurements where consent and licences allow. We do not release a voice beyond what its speaker agreed to, data whose licence forbids it, or a model before it passes Namu’s release checks. When something stays closed, we say what it is and why.
PeopleThe Lab is led by Namu’s founders
Mouhamad Mamane
Mouhamad Mamane co-founded Namu and leads it as chief executive. He and Abdourahamane Ide Salifou approve what the Lab publishes, including results and decisions about voices and data.
Abdourahamane Ide Salifou
Abdourahamane Ide Salifou co-founded Namu and leads its speech recognition research. In 2025 his entry won the AI for Good Swahili ASR Challenge on Zindi, with a 17.81% word error rate on the competition’s private leaderboard. He trained it on a single borrowed 16 GB GPU, and the code is public.
Help build speech AI for the languages you speak
Each programme page says what help it needs. Today we are looking for paid Hausa speakers for two-person recordings, native listeners and transcribers across Hausa dialects, Zarma writers and linguists, and researchers working on turn-taking and measurement. Write to contact@namuai.org and name the programme you want to work on.
Get in touch