I am a CS Researcher at University of Southern California, working under the supervision of professor Erdem Bıyık. My overarching research goal is to develop data-efficient, generalizable, and trustworthy robot learning frameworks that can operate robustly in complex, sequential settings. A central theme of my research vision is introducing structure into learning-based robotics. Rather than relying purely on end-to-end data-driven policies, I am interested in leveraging additional sources of guidance, such as prior knowledge, auxiliary modalities, and richer supervision signals, to support enhanced data curation, representation learning, policy optimization, and inference-time reasoning.
