At Hello Tractor, I checked source data for missing fields, inconsistencies, and formatting errors before loading, and verified model inputs and outputs against expected formats. I also wrote data-handling documentation used by the engineering and analytics teams.
I built CoffeeGuard, a production computer vision system for coffee leaf disease diagnosis. Its MobileNetV2 classifier was trained across 58,549 images and five disease classes, and achieved 98.80% accuracy and 98.36% macro F1; I also engineered FastAPI endpoints, an analytics dashboard, and retraining audit logs.
On Synapse Platform, I built a prompt generation layer for Google Gemini that adapts prompts to learners’ live assessment results, and implemented resource caching for generated material. I’m a native Swahili speaker from Kenya with fluent English and a Software Engineering degree specializing in Machine Learning.
