Junjie Liang

I work on AI security. At ByteDance, I build automated red-teaming systems for large language models and AI agents.

Before that, I spent three years at Bloomberg AI working on information retrieval. I earned my Ph.D. at Penn State, advised by Dr. Vasant G. Honavar in the AI Lab, where I worked on machine learning for longitudinal data. Before my Ph.D., I worked with Dr. Jinlong Hu at South China University of Technology during my master’s degree.

Research interests: AI security (LLMs and agents), information retrieval, natural language understanding, longitudinal data analysis.

Outside work, I snowboard, play badminton, lift weights, and enjoy math and RTS games; my all-time favorites are AOE2 and Don't Starve Together. I also build small open-source tools that I use every day, such as Marmota, a terminal file manager that works beside coding agents (Take the tour), and MarmotIM, a Wubi–Pinyin input method for macOS.