About
Uyir Lab builds AI that speaks the languages big models miss.
Uyir means life in Tamil. We are an independent AI lab making AI agents accurate in the languages big models get wrong, the way those languages are actually spoken. Big models learn from the written web, and most of the world does not talk the way the web is written.
What we build
Language Brain
A layer any AI agent plugs into. It understands dialects, slang and cultural context, so the agent you already have becomes accurate in languages it used to get wrong.
How it works →Nila
Our first model: a Tamil language model trained from scratch on Tamil, English and Tanglish, built for how Tamil is really spoken from Jaffna to Chennai.
Nila and Tamil →The contribution studio
Native speakers record their voice and share writing, with explicit consent. Every contribution is reviewed by people before it is used.
Give your voice →How we work
- Consent first
- We only train on data people chose to give us, or data that is licensed for commercial use.
- People review everything
- Every recording and text is checked by a person before it reaches a model.
- Depth over breadth
- One language at a time, every layer of it: dialects, slang, voice and culture.
- Communities own the story
- We work with native speakers, diaspora groups and local researchers, not around them.
Languages
Tamil is live. These are next on our roadmap, 16 languages in all.
- Tamil
- Sinhala
- Punjabi
- Swahili
- Hausa
- Amharic
- Javanese
- Fula
- Yoruba
- Bhojpuri
- Oromo
- Pashto
- Igbo
- Khmer
- Haitian Creole
- Quechua
Work with us
Building an agent for people we don't serve well yet? Or want to help with your language?