Bethlehem, Pennsylvania, United States
Production AI is hard to get right, but watching mental health professionals use a system built from scratch to practice real counseling scenarios made every late night worth it.
Under the supervision of Prof. Mooi Choo Chuah, this research centered on an LLM-powered training simulator built end-to-end — from prompt engineering and full-stack development to IRB-approved user studies. Reduced AI hallucinations by 35% through iterative prompt architecture, and validated a cultural conditioning approach with participants — 83% reported stronger emotional connection compared to baseline AI.
Findings on large language model cultural adaptation were presented at ABRCMS 2025, one of the largest multicultural research conferences in the country. This work redefined what it means to build AI responsibly, not just technically sound, but human-centered at every layer.