About C-elo Labs
We are pioneering AI for low-resource African languages, starting with Kikuyu and expanding to Kamba, Luo, and beyond.
Our Mission
Over 2,000 African languages are spoken across the continent, yet virtually none have meaningful AI support. At C-elo Labs, we believe language should never be a barrier to technology. We build production text and voice AI that speaks African languages natively.
Our approach: fine-tune state-of-the-art open-source models on curated African language datasets and deploy them as accessible, real-world products — starting with Kikuyu.
Our Focus
Neural Machine Translation
Fine-tuning models like Google's TranslateGemma-12B for English↔Kikuyu translation using LoRA. Our deployed model achieves 19.61 BLEU — a 758% improvement over zero-shot.
Voice AI Systems
Building Speech-to-Speech models using the Mimi neural codec adapted for Kikuyu tonal fidelity. Currently running a cascaded voice pipeline (ASR → LLM → TTS) with end-to-end streaming in development.
Dataset Engineering
We use the African Next Voices corpus (750+ hours of Kikuyu audio), Google's WAXAL TTS data, and 30,430 curated English-Kikuyu translation pairs to train our models.
Our Values
Language First
We prioritize linguistic authenticity, building AI that handles tonal nuances, agglutinative morphology, and code-switching.
Open Research
We publish our methods, share datasets, and open-source tools to accelerate progress across all low-resource languages.
Real-World Impact
We measure success by deployment—farmers getting advice in African languages, healthcare workers using voice diagnostics, students learning in their mother tongue.
Founders
Mark Gatere
Founder & Lead Researcher
Software engineer and ML practitioner based in Nairobi. Native Kikuyu speaker. Open-source contributor to vLLM and HuggingFace Transformers.
Get in Touch
We're seeking partnerships with universities, cooperatives, and investors who share our vision for African language AI. Reach out at hello@c-elo.com