Mhoro!
Shona for “hello”. Avail Labs · Applied AI research

AI that speaks our languages.

Avail Labs is our applied research arm. We build open datasets and language models for African languages, starting with Shona, and hand them to our sector teams to deploy.
Partner with Avail Labs
Read our data principles
Flagship project · In progress

Taura

/ta-u-ra/ · Shona, “to speak” · working title
A Shona-first language model for question answering, translation and voice, focused on agriculture and health.
Millions of people speak Shona, but most AI tools barely understand it. Taura is built to change that, with the people who speak it.
Our approach
  • 1

    Collect with consent

    Partner with Shona speakers, farmers, health workers and universities. Contributors give consent and are fairly paid.
  • 2

    Adapt, don’t start from zero

    Adapt strong open multilingual models with high-quality Shona data rather than training from scratch.
  • 3

    Evaluate with native speakers

    Native speakers review accuracy, tone and usefulness, not just benchmarks.
  • 4

    Release and deploy

    Release openly where consent allows; pilot inside Avail Agri and Avail Health.
Roadmap

From data to deployment.

  • Q4 2026 · now

    Data collection plan

    Partners, consent process, collection protocol, compute grants.
  • H1 2027

    Shona corpus v1

    Text and speech focused on agriculture and health.
  • 2027

    First fine-tuned model

    Taura v0.1 and an open evaluation set.
  • Next

    Pilot in Avail Agri and Avail Health

    Live with farmers and health workers, measured on the Availability Index.
Principles

How we do research.

  • Open by default

    Datasets, evaluations and models are released under open licences wherever consent allows.

  • Consent and fair pay for data contributors

    Every contributor knows how their words are used, and is paid for their time.

  • Community evaluation

    Native speakers decide if the model is good enough — not just a leaderboard.

  • Data sovereignty

    Data is stored and processed in Africa wherever possible.
For funders & compute partners

Help us teach AI to speak Shona.

We're seeking compute partners — if you can support African-language AI research, talk to us.
Email the founder
Research partners: Open to universities and research groups in Zimbabwe and across Africa.
  • GPU / TPU compute grants
  • University research collaborations
  • Data partnerships with farmer and health organisations
  • Native-speaker reviewers