At Welo Data, CrowdGen, and Remotasks, I evaluate and rank LLM outputs in Swahili, Kikuyu, and English for truthfulness, semantic accuracy, and cultural relevance. I maintain a 98%+ quality assurance score across major project cycles.
I’ve transcribed and timestamped 200+ hours of multi-dialect audio at 98%+ accuracy, capturing English-Swahili and Kikuyu-English code-switching and Sheng. I also audit AI-generated text, audio, and video descriptions, and annotate video with object tracking and bounding-box labels.
On the Kakamega Soko Mobile Marketplace Application, I designed and built an Android marketplace app, covering product scoping, Firebase backend database design, and iterative releases. I bring a focus on helping AI models sound accurate, natural, and culturally appropriate for East African users.

