Tony

Tony (Lipeng) He is a researcher and software engineer pursuing a Master of Mathematics in Computer Science at the University of Waterloo. His work focuses on trustworthy machine learning, particularly the adversarial robustness of AI systems and the security and privacy of agentic AI. He develops effective adversarial attacks and principled defenses, drawing on applied cryptography, theoretical ML, and computer security.

• Research on Adversarial Robustness: Studies how robustness breaks under realistic deployment conditions and develops methods to fix it, including feature-locking techniques for language models.
• Publications in Top Venues: Has published on LLM safety, adversarial attacks, and defenses at conferences like USENIX Security and ACM CCS.
• Industry Experience: Co-founded Bluelet AI as CTO, worked as a full-stack engineer at BioRender, and held software engineering roles at Safyre Labs and Bitbuy.
• Teaching and Mentoring: Serves as a teaching assistant and instructional apprentice at UWaterloo, and mentors in the Directed Reading Program on AI safety.
• Awards and Funding: Received multiple scholarships including the David R. Cheriton Graduate Scholarship and Lambda Research Grant.