Making AI work in Somali.
We develop open Somali language resources and work with partners on applications that help people find information, communicate and use services in Somali.
- Researching Maxaa Tiri and Maay
- Text, speech and archives
- Responsible, open research
- Serving public, private and NGO sectors
Serving Somali speakers wherever they are
From language data to useful tools
We start with the language as people use it: regional speech, specialist terms and knowledge held in local archives.
Data and Digital Archives
Somali text and speech datasets, archive digitisation and OCR, with cleaned and annotated corpora.
Explore data & archivesTranslation and Terminology
Somali translation and localisation with human review, shared glossaries and translation memories.
Explore translationSpeech and Audio Tools
Research and development for transcription, captions and text-to-speech across regional speech and recording conditions.
Explore speech toolsModels and Knowledge Tools
Model adaptation for Somali, plus document search and assistants designed to answer from approved sources.
Explore model developmentEvaluation and Human Review
Somali benchmarks and expert review of translation, reasoning, accuracy and safety.
Explore evaluationData Governance and Deployment
Consent, personal-data review and licensing, with hosting advice and training for your team.
Explore governanceStart with the words.
Keep the context.
- Listen & collect
Text, speech and knowledge, with permission.
- Review & document
Meaning, language variety and the limits of the data.
- Build & evaluate
Useful tools, assessed with Somali speakers.
Research grounded in Somali
Somali word structure, spelling and grammar shape our model work. We combine linguistic research with practical development.
-
Maxaa Tiri and Maay — separate data and testing plans for each, with speakers involved throughout.
-
Built for real conditions — a focus on efficient tools for limited computing power or connectivity.
-
Open and accountable — release plans that include clear licences, documentation and evaluation.
Start small, then grow
A collaboration can begin with a dataset, a model assessment or a defined application.
Define
We agree the goal: a dataset, a model assessment or an application.
Agree criteria
Together we set clear evaluation criteria before any work begins.
Pilot
We test a small pilot and review the results with your team.
Extend & hand over
You receive documented code, evaluation methods and support for hosting and maintenance.
Connect with Somali speakers
Start with a specific audience and a practical problem.
Working alongside Somali institutions
Linguists, researchers and students from our partner institutions contribute to language review, joint research and hands-on training.
Work with Buuni
Contact us about a research partnership or a Somali language technology project.