Elewa
UnderstandModels trained on how people really write and talk: code-switching, Sheng, slang and mixed scripts included.
AkiliNeuro trains and runs models that understand, chat and speak East African languages, starting with Kenya, Uganda and Tanzania, built with the people who speak them.
Habari!
Nzuri sana! Nikusaidie vipi leo?
Hello! / Very well! How can I help you today?
Models trained on how people really write and talk: code-switching, Sheng, slang and mixed scripts included.
Assistants that reply in the language you wrote in, with the local context that makes an answer useful.
Speech recognition and natural voices, so people who prefer talking to typing are not left out.
Translation between African languages directly, instead of routing every sentence through English.
We focus on East Africa first and go deep before we go wide. The first wave is the languages we hear every day. The rest follow as we build data partnerships. This list will change as we go.
No wrappers around someone else's model. Data, training and serving stay in our hands, so the languages get the attention they need.
We record and collect text and speech with native speakers, with consent, and we pay contributors for their work.
We train our own models on that data, tuned for the way each language is actually used.
We serve the models from our own infrastructure, built to stay fast on slow connections and cheaper phones.
You chat with them at akilineuro.ai. Developers will get an API to build on them.
Talk or type to Akili in your own language. Ask questions, draft messages, translate and learn.
Open akilineuro.ai →Put our language, speech and translation models inside your app, call centre or service.
Ask for early accessA language model is only as good as the people behind its data. We work with speakers, pay them fairly and credit their work.
Become a contributor →Read, record and review. Your voice helps the models sound like home.
Help us get grammar, tone and spelling right, language by language.
Universities, media houses and NGOs with text, audio or a use case.