MCAAI is proud to co-host the Afrivoice EAC Automated Speech Recognition Hackathon — a regional competition bringing together AI researchers, speech technologists, and language community representatives from across the East African Community to advance the state of ASR for the region's most under-resourced languages.
About the Hackathon
The Afrivoice EAC ASR Hackathon is a focused, results-driven competition designed to push the boundaries of automatic speech recognition for low-resource East African languages. Participants will be challenged to build, fine-tune, and evaluate ASR models on curated speech datasets drawn from MCAAI's own corpora — including the DhoNam Dholuo corpus and the African Next Voices multilingual dataset — as well as newly released data from partner institutions across Kenya, Uganda, Tanzania, Rwanda, and Burundi.
Unlike a traditional conference, the hackathon is structured around hands-on model building. Teams will have access to compute resources, baseline model checkpoints, and dedicated mentorship from MCAAI's NLP and speech engineering team throughout the competition period. The goal is not only to produce winning models, but to generate open-source tools and fine-tuned checkpoints that the broader East African AI community can build on after the event.
Target Languages
The hackathon focuses on ASR development for languages that are chronically under-served by mainstream speech technology platforms. The competition dataset covers the following language tracks:
- Dholuo — Nilotic language, Western Kenya (4M+ speakers)
- Kikuyu — Bantu language, Central Kenya (8M+ speakers)
- Luganda — Bantu language, Uganda (6M+ speakers)
- Kinyarwanda — Bantu language, Rwanda (12M+ speakers)
- Kirundi — Bantu language, Burundi (9M+ speakers)
- Open Track — Any additional EAC indigenous language with submitted data
Teams may compete in one or more language tracks. Cross-lingual and multilingual model submissions are explicitly encouraged and will be evaluated in a dedicated multilingual category.
Competition Structure
The hackathon runs over 72 hours of active model development, preceded by a half-day orientation session covering the dataset formats, evaluation metrics, and baseline model infrastructure. The competition is structured in two phases:
- Phase 1 — Public Leaderboard (Days 1–2): Teams submit models evaluated against a public development set. Word Error Rate (WER) scores are updated on a live leaderboard accessible to all participants.
- Phase 2 — Final Evaluation (Day 3): Teams make a final submission evaluated against a held-out test set. Final rankings are determined by WER on the hidden test data, with secondary scoring for model size efficiency and open-source code quality.
Prizes & Recognition
- 1st Place (per language track): KES 150,000 + invitation to present at IndabaX Kenya 2026
- 2nd Place (per language track): KES 75,000
- 3rd Place (per language track): KES 40,000
- Best Multilingual Model: KES 200,000 + MCAAI Research Collaboration Award
- Best Open-Source Contribution: Special recognition award + feature in MCAAI's annual research report
All submitted models that meet a minimum quality threshold will be considered for integration into MCAAI's open model repository, with full attribution to the contributing team.
Who Should Participate
The hackathon is open to individuals and teams of up to four members. We welcome participants from all backgrounds — university students, independent researchers, professional engineers, and community linguists. No prior experience with speech recognition is required; orientation materials and baseline code will be provided to help new participants get up to speed quickly. Teams with a mix of technical and linguistic expertise are strongly encouraged, as domain knowledge of target languages is a genuine competitive advantage.
Key Dates
- Registration Opens: 1 July 2026
- Registration Deadline: 10 August 2026
- Dataset Release: 15 August 2026
- Hackathon Dates: 5–7 September 2026, Maseno University
- Winners Announced: 7 September 2026 (closing ceremony)
To register your team or find out more, visit the MCAAI events page or contact the hackathon organising team at afrivoice@mcaai.ac.ke.
