At Kenya Revenue Authority, I annotated and quality-checked text and audio datasets, then evaluated AI responses for accuracy, relevance, tone, safety, and instruction-following. I also identified hallucinations and factual errors and shared structured feedback to support model fine-tuning.
I audited recordings against transcripts and moderated user- and AI-generated content against safety guidelines. When project instructions were ambiguous, I documented the issue and created escalation pathways for edge cases.
