Mansi Phute
My research interests are Responsible AI and ML safety.
I work on developing explanations for ML systems, analyzing them to identify vulnerabilities, and finding solutions to mitigate these issues.
My UNDREAM system system offers a way to bridge differentiable rendering and photorealistic simulation for end-to-end adversarial attacks, thus enabling beter transferability of attacks to the physical world.
My work includes LLM Self Defense, which leverages the model's own understanding of harm to protect itself from attacks.
Education
Summer 2024 —
Ph.D. in Computer Science
Georgia Institute of Technology, Atlanta, GA
Fall 2022 — Spring 2024
M.S. in Computer Science
Georgia Institute of Technology, Atlanta, GA
Specialization: Machine Learning
Fall 2018 — Spring 2022
B.Tech. in Electronics and Telecommunication
Honors: Artificial Intelligence and Data Analytics
Research Experience
Summer 2022 — Present
Georgia Institute of Technology, Atlanta, GA
Graduate Research Assistant School of Computational Science and Engineering
Advisor:
Duen Horng (Polo) Chau
Member of the Polo Club of Data Science where we bridge and innovate at the intersection of data mining and human-computer interaction to synthesize scalable, interactive, and interpretable tools that amplify human’s ability to understand and interact with big data. Developed defences against adversarial attacks in Language and Vision domain.
Spring 2023
Georgia Institute of Technology, Atlanta, GA
Graduate Teaching Assistant School of Computational Science and Engineering
Mentor:
Duen Horng (Polo) Chau
Fall 2021 — Spring 2022
Nanyang Technological University, Singapore
Undergraduate Research Assistant Cyber Security Research Centre at NTU (CYSREN)
Mentor:
Thambipillai Srikanthan
Increasing python application security by analyzing libraries used. Developed dynamic dependency graph to trace vulnerabilities. Automated human resource planning and forecasting by combining business intelligence of NHS, UK with data analytics to properly shift the HR planning from manual to automated.
Spring 2021
Undergraduate Research Assistant Associated with Dassault Systems
Mentor:
Jyoti Madake
Developed AI based solutions for agricultural problems faced in India by usimg hyperspectral imaging to predict soil fertility in the land
Fall 2020
Undergraduate Research Assistant School of Electronics and Telecommunication
Mentor:
Abha Marathe
Conducted a thorough literature survey on the use of AI in finance and the various ways it is used for risk management in the stock market
Industry Research Experience
Summer 2025 — Fall 2025 Present
HiddenLayer, Inc., Austin, TX
Research Assistant Adversarial Robustness Team
Mentor:
Jason Martin,
Ravi Balakrishnan
Helped pioneer transition of AI defense systems to account for multimodal attacks. Developed universal transferrable multimodal steering images that can alter model behavior using the input channel without requiring access to the model internals. My work during the internship was implemented into the product [AIDR (AI Detection and Response)](https://www.hiddenlayer.com/aidr/)
Summer 2019
Tech Mahindra Ltd, Pune, India
Mentor:
Rahul Bedmutha
Developed a portal for internal use, using HTML, CSS and Javascript.
Honors and Awards
2024
Marshall D. Williamson Fellowship
Awarded to a well-rounded, second-year Master's student who best embodies values of academic excellence and leadership
2022
Best Scholar in ECE
Merit-based award for the ECE student with the highest undergraduate GPA in the entire department
Publications
arXiv. 2025.
Ravi Balakrishnan,
Mansi Phute
arXiV. 2025.
Mansi Phute,
Ravi Balakrishnan
arXiv. 2025.
Matthew Hull,
Haoyang Yang,
Pratham Mehta,
Mansi Phute,
Aeree Cho,
Haoran Wang,
Matthew Lau,
Wenke Lee,
Willian Lunardi,
Martin Andreoni,
Duen Horng Chau
arXiv (arXiv). 2025.
Matthew Hull,
Haoyang Yang,
Pratham Mehta,
Mansi Phute,
Aeree Cho,
Haoran Wang,
Matthew Lau,
Wenke Lee,
Willian Lunardi,
Martin Andreoni,
Duen Horng Chau
CVPR Workshop on Neural Fields Beyond Conventional Cameras (NFBCC) (CVPR'25). 2025.
Main, Conference on Empirical Methods in Natural Language Processing (EMNLP). 2025.
Matthew Hull,
Haoran Wang,
Matthew Lau,
Alec Helbling,
Mansi Phute,
Chao Zhang,
Zsolt Kira,
Willian Lunardi,
Martin Andreoni,
Wenke Lee,
Duen Horng Chau
International Joint Conference on Artificial Intelligence (IJCAI) (IJCAI'25). 2025.
Anisha Pal,
Julia Kruk,
Mansi Phute,
Manognya Bhattaram,
Diyi Yang,
Duen Horng (Polo) Chau,
Judy Hoffman
NeurIPS. 2024.
Seongmin Lee,
Zijie J. Wang,
Aishwarya Chakravarthy,
Alec Helbling,
ShengYun Peng,
Mansi Phute,
Duen Horng (Polo) Chau,,
Minsuk Kahng
ACL demo. 2024.
ShengYun Peng,
Weilin Xu,
Cory Cornelius,
Matthew Hull,
Kevin Li,
Rahul Duggal,
Mansi Phute,
Duen Horng (Polo) Chau,
Jason Martin
BMVC. 2023.
Mansi Phute,
Alec Helbling,
Matthew Hull,
ShengYun Peng,
Sebastian Szyller,
Cory Cornelius,
Duen Horng (Polo) Chau
ICLR Tiny Paper. 2024.
Talks and Presentations
Large Language Model Evaluation
March 2025
Georgia Institute of Technology, CS 8001: Large Language Models
Large Language Models and How They Work
June 2024
Georgia Institute of Technology, CS 8001: Large Language Models
LLM Self Defense: By Self Examinations, LLMs Know They Are Being Tricked!
October 2023
IBM, San Jose CA
Press
April 2024
"Student Excellence Honored at Annual Event," Georgia Tech
August 2023
May 2020
"Team Eklavya- E&TC; students team Designs Autonomous sanitisation robot," Vishwakarma Institute of Technology
Teaching
Spring 2023
Georgia Institute of Technology, Atlanta, GA
I was a Teaching Assistant (TA) at Georgia Tech for the class Data and Visual Analytics where I worked with a team of 30 TAs to enable learning in a class of more than 1200 students. I was a part of designing homework and mentoring students in their course work and project work.
Fall 2019
Vishwakarma Institute of Technology, Pune, India
A semester long teaching program where I created learning opportunities for increasing literacy in society aimed towards people outside the traditional schooling age. Thus proving that there is no binding of age to learn how to read or write. This program aimed at combating illiteracy in specific sections of society.
Service
Reviewer
NeurIPS Workshop on Socially Responsible Language Modelling
(NeurIPS SoLaR)
2023
Mentoring
B.S. in Computer Science, Georgia Institute of Technology
References
Dr. Polo Chau, Associate Professor
School of Computational Science and Engineering
Georgia Institute of Technology
Dr. Thambipillai Srikanthan, Professor
School of Computer Science and Engineering
Nanyang Institute of Technology