Kelvin Yuxiang Huang

I am a first-year Ph.D. student advised by Prof. Gururaj Saileshwar in the SITH Lab.

Prior to my Ph.D., I completed my undergraduate studies in Computer Science and Statistics at the University of California, Berkeley.

My research primarily focuses on agentic system security, with an emphasis on mitigating system-level security risks in autonomous and tool-using AI agents. I am particularly interested in runtime security, privilege-boundary enforcement, attack-path abstractions, and security monitoring for agentic systems.

Previously, my research spanned trustworthy AI and adversarial robustness. At UC Berkeley, I conducted research on trustworthy agentic AI advised by Prof. Dawn Song, focusing on structured observability for agent systems. I also worked on adversarial robustness for large vision-language models, advised by Prof. Qingyun Wang.

Earlier, I also worked on 3D generation at HKU with Prof. Xihui Liu, and on computational vision at UC Berkeley with Prof. Dennis Levi.