Mingkai Li

About Me

I am a 2nd-year Ph.D. student at Columbia University, advised by Prof. Tanvir Ahmed Khan, and working closely with Prof. Suman Jana and Prof. Joe Devietti.

My research centers on ensuring that complex, AI-integrated computing systems are correct, reliable, and secure — especially as such systems become critical infrastructure yet remain difficult to fully reason about.

I am driven by two interconnected questions:

• System Assurance: Bugs in ML systems can emerge from the model, runtime, hardware, or their interactions — making reliability guarantees hard to establish. I am interested in techniques that systematically surface these failures before they cause real harm.
• Security By Design: LLMs and agentic systems are fundamentally shifting the attacker-defender balance. I am interested in how the security community should respond through principled, proactive defenses rather than purely reactive ones.

Prior to Columbia, I received my B.E. in Computer Science from University of Science and Technology of China (USTC). I have had the pleasure of working with Prof. Prateek Saxena and Prof. Trevor E. Carlson at National University of Singapore, and Prof. Xianghang Mi at USTC. Outside of research, I enjoy cycling, squash, and swimming.

Feel free to reach out if you share similar interests!

News

• [Mar. 2026] Our paper NanoTag is accepted to IEEE S&P 2026.

Publications

Workshops

Talks

• NanoTag: Systems Support for Efficient Byte-Granular Overflow Detection on ARM MTE
  • IEEE S&P 2026, May 2026, San Francisco, CA
  • NY Architecture & Systems Day 2026, May 2026, Princeton University, NJ
• Challenges and Design Considerations for Finding CUDA Bugs Through GPU-Native Fuzzing
  • HotEthics 2026, March 2026, Pittsburgh, PA

Teaching

• Teaching Assistant, EECS6894: Data Center Processing, Prof. Tanvir Ahmed Khan, Fall 2025
• Teaching Assistant, CSEE4119: Computer Networks, Prof. Ethan Katz-Bassett, Fall 2024

Powered by Jekyll and Minimal Light theme.