At AUB / AUB Medical Center, I developed and validated clinical machine-learning models to predict amblyopia treatment response and final visual outcomes. The models achieved ROC-AUC scores of 0.75 for treatment response and 0.92 for final vision, and I contributed to a study accepted for publication in the Journal of AAPOS.
For my master’s thesis, I developed an uncertainty-aware hybrid ML–LLM decision-support framework and built and deployed a FastAPI research prototype. Its agreement-based gating achieved 72.8% accuracy at 69.5% coverage.
I also collaborated on Haris, a LangGraph security layer for multi-agent AI, and led the ResQDrones project, integrating YOLOv8 fire detection, embedded sensors, and AWS IoT communication. My research includes evaluating LLM robustness to health misinformation.
