I designed and ran a 15-day safety audit of an AI mental-health companion, evaluating whether it gathered decision-relevant context before offering advice or risk-related guidance. I delivered 18 severity-ranked findings with pass tests and evidence, and the founder shipped fixes before the audit closed.
At Cognishine, I evaluated AI-generated text, image, and video for accuracy, tone, relevance, and alignment with user intent, and reviewed psychology content against evidence-based guidelines. My clinical psychology training includes supervised intake interviewing, behavioral observation, case formulation, and therapy; I also researched emotion recognition using PsychoPy and SPSS.
