AI Alignment & Technical Subject Matter Expert (Safaricom PLC)
Built and optimized gold-standard datasets to improve multi-step fintech conversational performance using RLHF methods. Audited outputs for hallucination mitigation and accuracy in financial reasoning and technical documentation contexts. Supported downstream model refinement by turning evaluation findings into higher-quality training examples. • Created “Gold Standard” Chain-of-Thought datasets for multi-step fintech queries. • Applied RLHF to improve conversational accuracy for M-Pesa AI customer interfaces. • Conducted security and hallucination-focused auditing of Python and SQL code snippets. • Performed red-teaming to identify logic failures for targeted dataset/model fixes.