Domain expert contributor for AI training, evaluation, and alignment
As a domain expert contributor, I specialized in ranking and evaluating model outputs related to psychiatric, medical, and clinical-reasoning tasks. I participated in adversarial red-teaming and safety alignment, including high-risk mental health content like suicidality and substance use. My responsibilities included hallucination audits, reasoning step verification, creation of clinical case vignettes, and multilingual model evaluation, especially in Kiswahili and Kikuyu. • Evaluated and rated model outputs for clinical accuracy using DSM-5-TR and ICD-11 criteria • Conducted red-teaming for safety testing on sensitive mental health scenarios • Authored benchmark problems for psychiatric diagnostic reasoning and treatment planning • Performed multilingual clinical quality assessments in English, Kiswahili, and Kikuyu