Kevin Dai

Picture of Kevin Dai

Hi! I’m Kevin, and I am currently a second-year PhD student in Computer Science at Georgia Tech, where I work on the implications of machine learning for security and privacy under the guidance of Professor Teodora Baluta in the School of Cybersecurity and Privacy. Previously, I graduated from Amherst College summa cum laude with a degree in math, computer science, and statistics. My senior thesis, with Professor Ryan Alvarado in the Department of Mathematics, examined the foundations of reproducing kernel Hilbert spaces. In my free time, I enjoy rock climbing, the performing arts, and exploring new restaurants in Atlanta.

My current research interests focus on understanding generative AI from the lens of security and privacy. Broadly speaking, this means exploring potential limitations in modern AI systems that attackers are able to effectively exploit. Right now, I am working on potential vulnerabilities in embedding models that may cause information leakage. I am also working on a human survey for views on privacy through the framework of contextual integrity, which posits that privacy depends on the context in which information is shared. This builds upon my previous work that examined a similar question in large language models, which, in an ideal world, would reflect human values.

Selected Publications, Presentations, and Posters

  • Dai, K. T., Baluta, T., Challenges in Evaluating Contextual Integrity for Large Language Models. In submission
  • Dai, K. T., Gisolfi, N., Miller, J. K., Dubrawski, A., Distributed Testing for Robust Inference. Robotics Institute Summer Scholars Working Papers Journal. 2024.
  • The Impact of ChatGPT on Consultants. Artificial Intelligence in the Liberal Arts Undergraduate Conference. April 2024.
  • The Impact of ChatGPT on Consultants. MIT Undergraduate Research Technology Conference. October 2023.
  • Is It Pop?: Music Genre Classification. ACM Massachusetts Gender-Inclusive Computing Celebration. April 2023.
  • Wu, X., Markir, A., Xu, Y., Hu, E. C., Dai, K. T., Zhang, C., Shin, W., Leonard, D. P., Kim, K., Ji, X., Rechargeable Iron–Sulfur Battery without Polysulfide Shuttling. Adv. Energy Mater. 2019, 9, 1902422. https://doi.org/10.1002/aenm.201902422

Contact

  • Office: CODA 9th floor, 756 W Peachtree St NW, Atlanta, GA 30308