Kuntai CAI

caikt@comp.nus.edu.sg

I am Kuntai Cai, a research scientist at TikTok Singapore, with a focus on AI agent and privacy protection. My academic journey began with a Bachelor’s degree in Computer Science and Technology from Tsinghua University, which I earned in 2019. My PhD focuses on differential privacy and data synthesis, conducted under the guidance of Professor Xiaokui Xiao at the National University of Singapore.

In June 2021, my supervisor, collaborators, and I won 1st Place in the NIST Differential Privacy Temporal Map Challenge. Our winning solution was an innovative adaptation of our previous work, PrivMRF, which we have open-sourced as a foundational tool for privacy protection. More recently, at SIGMOD 2023, my collaborators and I introduced PrivLava, the first solution for synthesizing relational data with foreign keys while ensuring differential privacy.

I am deeply committed to advancing research in AI agent and privacy protection. My work seeks to bridge theoretical innovation with practical applications, driving progress in both academia and industry.

Research Interests

  • AI for Database Testing
  • Differential Privacy
  • Data Synthesis
  • Graphical Models

Professional Service

  • 2024 & 2025 Reviewer, IEEE Transactions on Knowledge and Data Engineering
  • 2022 Reviewer, ACM Transactions on Database Systems

Publications

2025

  1. PrivPetal: Relational Data Synthesis via Permutation Relations
    Kuntai Cai, Xiaokui Xiao, and Yin Yang
    Proceedings of the ACM on Management of Data, 2025
  2. GCON: Differentially Private Graph Convolutional Network via Objective Perturbation
    Jianxin Wei, Yizheng Zhu, Xiaokui Xiao, Ergute Bao, Yin Yang, Kuntai Cai, and Beng Chin Ooi
    In 2025 IEEE International Conference on Data Engineering, 2025

2023

  1. PrivLava: Synthesizing Relational Data with Foreign Keys under Differential Privacy
    Kuntai Cai, Xiaokui Xiao, and Graham Cormode
    Proceedings of the ACM on Management of Data, 2023

2021

  1. Data Synthesis via Differentially Private Markov Random Field
    Kuntai Cai, Xiaoyu Lei, Jianxin Wei, and Xiaokui Xiao
    Proceedings of the VLDB Endowment, 2021