Gefei Tan

gefei-upright.jpg

2233 Tech Drive

Evanston, IL 60208-3109

I am a fourth-year PhD candidate in the Computer Science Department at Northwestern University, fortunate to be advised by Xiao Wang and Ning Luo. From 2025 to 2026, I was also a Student Researcher in Google’s Private Computing group, mentored by Mariana Raykova.

My research focuses on the security, privacy, and verification of machine learning systems, particularly LLMs and autonomous AI agents. I combine cryptographic tools with machine learning to build rigorous, provable guarantees that remain robust in adversarial settings and efficient for production models.

I am a recipient of the Google PhD Fellowship in Privacy, Safety, and Security, and was selected for ACM Future Leaders of AI.

research interests

  • Verifiable machine learning: Designing proof systems and cryptographic protocols to verify model provenance, training integrity, and safety and privacy properties, without exposing proprietary weights or private training data.
  • Privacy-preserving machine learning: Co-designing cryptographic protocols with learning and alignment algorithms, so that models can be trained, fine-tuned, and repaired on sensitive data without exposing it.
  • LLM and agent security: Developing automated red teaming of LLMs and tool-using agents, using reinforcement learning to discover diverse jailbreak and prompt-injection attacks and to strengthen defenses against adaptive attackers.

More broadly, my work in applied cryptography spans secure multiparty computation and zero-knowledge proofs, with applications to secure databases and formal verification.