# Zhagaram Labs

> Zhagaram Labs builds speech and language models for Tamil first, for the more than 80 million people who speak it. Its first model is Kural, a voice model you can talk to in everyday Tamil. Most of what it builds is open source, and every claim comes with a number anyone can check on Uraikal, its open benchmark.

Every page on https://zhagaramlabs.com has a markdown version with the same content, listed below.

## Pages

- [Home](https://zhagaramlabs.com/index.md): what we build, the model roadmap, first benchmark results, and questions
- [Why Zhagaram Labs exists](https://zhagaramlabs.com/why.md): thirteen places today's AI fails Tamil speakers, the evidence for each, and what we're building
- [What makes Tamil hard for AI](https://zhagaramlabs.com/nuances.md): fourteen Tamil nuances, why AI gets each wrong, and what we're doing about it

## Code, models and data

- [Uraikal on GitHub](https://github.com/zhagaramlabs/uraikal): the open Tamil benchmark and toolkit, Apache 2.0
- [Uraikal on PyPI](https://pypi.org/project/uraikal/): `pip install uraikal`
- [Hugging Face](https://huggingface.co/zhagaramlabs): model weights and datasets
- [GitHub](https://github.com/zhagaramlabs): all code

## Contact

- Email: hello@zhagaramlabs.com