Guangzhi (Allen) Su

I’m Guangzhi (Allen) Su, an M.S. student in Artificial Intelligence and Innovation at Carnegie Mellon University, expected to graduate in May 2028. I graduated magna cum laude from Duke University in May 2026 with a B.S. in Applied Mathematics and Computer Science through the Duke–Duke Kunshan dual-degree program.

My work spans large language models, multimodal AI, multi-agent systems, and clinical AI, with a focus on safety, efficiency, and reliability. I currently conduct research with Prof. Rishi Kamaleswaran at Duke University School of Medicine on multimodal clinical risk prediction, and previously worked with Prof. Yiran Chen on efficient long-video understanding.

Alongside research, I build AI products as the founder of Prospectus.AI. I also served as the founding president of the DKU Computer Science Club and the DKU Finance Club, and as a student advisor for the DKU AI Club. In 2023, I was selected as a United Nations Millennium Fellow, working on education equity and cross-cultural exchange. Outside of work, I enjoy singing and sports; I spent four years as a tenor in the DKU Chorus and one year on the DKU badminton varsity team.

Research Interests

• AI Safety

  • Adversarial attacks and defenses for deep and multimodal models
  • AI-generated content detection and watermarking
  • Privacy risks and membership inference attacks

• Efficient AI Systems

  • Token and patch pruning for large models
  • Parallel and distributed training/inference
  • Mixture-of-experts (MoE) architectures for scalable deployment

• AI in Clinical and Healthcare Applications

  • Adverse event detection from clinical videos
  • AI-assisted decision making for clinicians
  • Fairness and reliability of models used in healthcare workflows