John Wangechi

I’m an AI Evaluation Specialist and Prompt Engineering professional with 5+ years of experience improving large language model performance through RLHF, model evaluation, data annotation, prompt optimization, content quality assessment, and AI safety review. I evaluate AI-generated outputs across conversational, educational, technical, and creative domains—ensuring factual accuracy, instruction adherence, and alignment with user intent. I take pride in translating complex evaluation criteria into clear, actionable feedback that raises model quality, reliability, and safety.

In my current role, I evaluate thousands of AI-generated responses against structured quality frameworks measuring accuracy, relevance, reasoning, safety, and instruction adherence. I strengthen evaluation consistency by applying RLHF methodologies to rank outputs and identify high-impact improvement opportunities, and I design prompt-testing scenarios to assess reasoning, creativity, factuality, and problem-solving performance across diverse use cases. I also identify hallucinations, logical inconsistencies, bias risks, and safety concerns, and I run red-team testing exercises to probe robustness against edge cases and adversarial prompts—backed by detailed reports and recommendations that inform model refinement.

Previously, I supported training and data quality through systematic AI response review, quality assurance across large annotation datasets, and assessments of linguistic, cultural, and contextual appropriateness for global usability. I contributed structured feedback and recommendations that improved training-data reliability and model performance while collaborating with distributed remote teams to meet milestones without sacrificing quality. Across evaluation, annotation, and safety work, my professional ethos is consistent: verify the right signals, surface risks early, and help teams ship AI that users can trust.