At Handshake AI, I evaluated AI model responses for accuracy, helpfulness, tone, and safety against detailed rubrics. I flagged ambiguous or inappropriate outputs and wrote concise rationales for my judgments.
At Silencio AI, I recorded English and Swahili speech and environmental audio for AI training datasets, checking clips for clarity, background noise, and correct labeling. In my Smart Crop Recommendation System project, I corrected mislabeled source data before training a Random Forest model that reached 99.5% test accuracy.
