Hi, I'm Mansi Phute

CS PhD at Georgia Tech

I am a PhD student at Georgia Tech advised by Polo Chau in the Polo Club of Data Science.

My research is in Trustworthy, Responsible, and Secure AI. I find vulnerabilities in AI systems and build practical, scalable defenses against them, so that large-scale models can be secured and the systems deployed with them can be trusted.

My work has led to two patents and two industry-deployed defenses, along with papers at top AI conferences. This includes VISOR and VISOR++: a single, universal steering image that steers architecturally diverse models without access to their internals at runtime.

I have collaborated with scientists at

IBM,

HiddenLayer,

Intel Labs, and

Nanyang Technological University

News

🎉 Oct 2026 I gave an invited talk at IBM : "Holistic Safety of Multimodal Multi-Agent Systems".
🎉 Sep 2026 I was featured in Georgia Tech College of Computing news: Internship Highlights.
🎉 Aug 2026 I completed my Research Internship at IBM, San Jose, CA.
🎉 Jun 2026 I presented UNDREAM at PAKDD 2026.
🎉 Apr 2026 My patent US12608861B1 was published.