Skip to main content Skip to navigation

Dr Kevin Han Huang

Website: kevinhanhuang.page

Email: kevin[dot]huang[at]warwick.ac.uk

Office: MB3.13


I am a postdoctoral research fellow funded by the EPSRC ProbAI Hub, working with Gareth Roberts at Warwick statistics and Boris Hanin at Princeton Operations Research & Financial Engineering. I received my PhD in machine learning from the Gatsby Computational Neuroscience Unit at UCL, where I was advised by Peter Orbanz at Gatsby and Morgane Austern at Harvard statistics. I was a visiting researcher at the LIPS group at Princeton Computer Science advised by Ryan P. Adams during Spring 2024, where I worked on AI-for-physics algorithms. Prior to my PhD, I received my Bachelor and Master in mathematics from University of Cambridge.

For the academic year 25-26, I am co-organising the ProbAI online seminar. I am also organising the ProbAI Theory of Scaling Laws Workshop at Warwick in summer 2026; stay tuned for more info.

URSS projects 2026

I am open to URSS project supervisions on one of the following topics:

  • ML models for tabular data with applications to AI safety;
  • Accelerating statistical algorithms with on-GPU parallel computing.

For enquiry, please email me with a CV and transcript by 15 Jan 2026.

I am a machine learning (ML) theorist, with increasingly frequent excursions to the applied side of things. On the theory end, I study the properties of large-scale stochastic systems that arise in ML and statistics. I develop and apply tools from:

  • Probability theory, e.g. universality and random matrix theory;
  • High-dimensional statistics, especially for non-linear estimators and dependent data;
  • Symmetry-based inference;
  • Stochastic optimisation and sampling theory.

As applications, I am primarily concerned with ML problems at scale with mathematical and statistical structures. Some examples include

  • Scaling laws of neural networks;
  • Algorithm design for AI for materials science;
  • Robustness and safety of AI models.

Let us know you agree to cookies