Ask most AI assistants a question in Kiswahili and you’ll usually get an answer. Ask in Gĩkũyũ, Dholuo or Luganda, or mix in some Sheng the way people in Nairobi actually talk, and things fall apart quickly. The replies are stiff, translated word for word, or quietly switch back to English.
That gap is why AkiliNeuro exists. Akili means intelligence or mind in Kiswahili. We want AI that understands, chats and speaks the languages people in East Africa use every day.
Why East Africa first
Africa has thousands of languages. Trying to cover them all at once would mean doing every one of them badly. So we’re starting at home:
- Kenya: Kiswahili, Sheng, Gĩkũyũ, Dholuo, Luluhya, Kalenjin, Kĩkamba, Ekegusii, Kĩmĩĩrũ, Maa, Somali and more
- Uganda: Luganda, Runyankore, Lusoga, Acholi, Ateso and more
- Tanzania: Kiswahili, Kisukuma, Kichagga, Kihaya, Kinyamwezi and more
Together that’s more than 200 languages. Many of them cross borders: Kiswahili, Dholuo, Maa and Ateso are all spoken in more than one country. Going deep in one region lets us get the details right, like spelling, tone, proverbs and code-switching, before we expand.
What we’re building
- Models that understand how people really write and talk, including mixing languages in the same sentence.
- A chat assistant at akilineuro.ai that replies in the language you wrote in.
- Speech, both recognition and natural voices, because many people would rather talk than type.
- Translation directly between East African languages, instead of routing everything through English.
We train these models ourselves and run them on our own infrastructure, so the languages get the attention they need.
Built with the people who speak them
A language model is only as good as its data, and that data comes from people. We work with speakers, linguists and partners. We get consent, pay contributors fairly and credit their work.
If you speak one of these languages and want to help, write to us at [email protected].
What’s next
The chat at akilineuro.ai is in beta. This blog is where we’ll share what we learn along the way: the hard parts of each language, how we collect data, and how the models are improving.
Karibu. Welcome.