At Mindrift AI, I evaluate AI-generated responses for relevance, accuracy, completeness, and instruction following. I compare and rank alternative responses, document my reasoning, and create and refine prompts to test model behavior.
I also label and classify data, check annotations for consistency, and investigate factual or logical errors. My work includes applying evaluation guidelines to ambiguous cases and recording findings for quality-control and model-improvement workflows.
As a Data Analytics and Revenue Intern at Nairobi city county, I reviewed business datasets and documents, flagged data-quality issues, and provided feedback on findings for AI training processes. I also worked with financial information under data privacy and security guidelines.
At Kenya Revenue Authority, I collected and verified tax data from more than 100 individuals and businesses, achieving 98% accuracy in revenue reporting. I analyzed income records for discrepancies, used Excel to track collection performance and non-compliance trends, and submitted field reports with findings and recommendations.
