Sunbird AI is a non-profit research organization based in Kampala, Uganda that builds and deploys practical AI systems for African contexts. Their focus areas include: speech/language processing, geospatial and environmental sensing, and public-sector decision support. Key technical work: ASR, TTS, translation, and language models for African languages; remote sensing and urban growth mapping; biodiversity acoustics; and tools for evidence-informed decision-making in public institutions. They maintain curated corpora, annotation guidelines, and evaluation sets as data infrastructure. Their flagship product is Sunflower, a multilingual LLM series (14B and 32B parameters) supporting 31+ Ugandan languages for translations, summaries, and explanations.
While Uganda is in East Africa (not WCA), Sunbird AI's approach and technology stack is highly relevant as a model for UNICEF's language technology initiatives in WCA. Their Sunflower LLM architecture could be adapted for WCA languages. Their experience with citizen feedback tools, environmental monitoring, and public-sector AI deployment provides valuable lessons. Could be a technical partner or advisor for similar initiatives in WCA countries.
Strong technical team with connections to Makerere University. Published 37 models on HuggingFace including the Sunflower LLM series with multiple quantization options. Maintains 7 datasets including SALT translation data and urban noise annotations. API available at api.sunbird.ai. Team of 12 active contributors on HuggingFace. Focus on open source and open research with commitment to locally-owned AI systems.