Crucial Mail

I improve AI language models through prompt evaluation, RLHF, output verification, and data annotation, grounded in research across pragmatics, semantics, and language use. At Mindrift, I conducted pairwise search-engine evaluations to improve recommendation accuracy, evaluated LLM outputs for hallucinations, safety, and factual alignment, and annotated video, image, and text datasets. I also achieved 29 approved domain-expert model-stumping submissions. I designed evaluation rubrics and authored more than 35 successful adversarial prompts to identify model failure modes, accuracy gaps, and logical inconsistencies. Alongside AI work, I taught pragmatics, semantics, and English grammar at Nakawa State University, mentored undergraduate research, and reviewed more than 30 academic manuscripts for MOPOLI Journal.