Kangyu's Homepage

Hello! I am Kangyu Zhu, a machine learning researcher working on the reliability and trustworthiness of multimodal large language models and agentic AI. I am currently at TikTok, where I build agentic AI systems for high-stakes real-world applications.

I received my Master of Science degree in Computer Science at Brown University. Before joining Brown in 2023, I earned dual degrees in Electrical Engineering from Zhejiang University and University of Illinois Urbana-Champaign.

Research Background

My research centers on reliable and trustworthy machine learning — the factuality, alignment, and robustness of large models in domains such as healthcare, where errors carry real cost.

I am currently most interested in:

• Multimodal Large Language Models — grounding, factuality, and preference alignment across vision and language.
• Agentic AI — reliable planning, tool use, and decision-making for LLM agents.
• Continual Learning & Agent Learning — how agents keep improving from experience, including in decentralized and privacy-preserving settings.

Collaborations

I’ve had the privilege of working closely with many distinguished researchers, including Prof. Timothy Miller at Boston Children’s Hospital and Harvard Medical School, Prof. Manling Li at Northwestern University, Prof. Huaxiu Yao at UNC-Chapel Hill, and Prof. Kang Li at the West China Biomedical Big Data Center. Many thanks to anyone who guided or helped me.