Practical work. Somali first.
Explore our areas of work. Every collaboration starts by agreeing the scope, available data and how results will be evaluated.
Research and project work. These areas describe collaboration opportunities. Availability and performance are established for each project. See resources and research status.
Somali Data and Digital Archives
Text and speech datasets built from how people really speak and write, with archives brought online.
- Data collection and annotation
- Archive digitisation and OCR
- Parallel translations and reviewed transcripts
Translation and Terminology
Somali translation and localisation that keeps meaning and technical terms consistent.
- Documents, websites and learning materials
- Human review by Somali linguists
- Shared glossaries and translation memories
Speech and Audio Tools
Develop and evaluate speech technology for regional accents, noisy recordings and mixed-language talk.
- Transcription and captions
- Text-to-speech
- Media archives, interviews and voice access
Models and Knowledge Tools
Adapt models for Somali and develop tools that answer questions from sources you approve.
- Continued pretraining and fine-tuning
- Somali tokenizers and linguistic tools
- Document search and AI assistants
Evaluation and Human Review
Know how well a model really works in Somali before you rely on it.
- Somali benchmarks for translation, reasoning and safety
- Expert review of factual accuracy
- Before-and-after model comparisons
Data Governance and Deployment
Responsible data from the start, and a system your own team can run.
- Consent, personal-data review and licensing
- Hosting and integration advice
- Documentation and team training
Not sure where to start?
Most projects begin with one dataset, one model assessment or one small pilot.