Jiayi Li

Teaching & mentoring

My teaching at UCLA has ranged from introductory statistics for life science students to preparation for a master’s in applied statistics and data science. In research, I have mentored undergraduate, master’s, and PhD students on questions in geometry and machine learning.

My priority in teaching and mentoring is to help students think independently: to formulate questions, examine assumptions, and justify their conclusions. I encourage them to explain their reasoning and consider alternatives, using discussion and feedback to help them develop their own judgment.

Research mentoring

Grokking and the geometry of neural networks

LOGML 2025 · Mentor
PhD students

I mentored a team of PhD students at the London Geometry and Machine Learning Summer School (LOGML) on the geometry and statistical complexity of neural networks.

This collaboration led to our 2026 preprint on grokking. Combining exact calculations of local model complexity with experiments in shallow quadratic networks, we studied why generalization can emerge long after a network fits its training data.

Estimating Bayesian complexity

LOGML 2026 · Co-mentor
PhD students

I co-mentored a PhD student team on estimating the local learning coefficient, a measure of local model complexity. The project asked whether variational inference could provide a more stable and scalable alternative to sampling-based methods, with an emphasis on implementation and empirical evaluation.

IPAM RIPS 2025. Academic mentor for an undergraduate research team studying the geometry of representations in large language models, through UCLA’s Research in Industrial Projects for Students programme with Microsoft Research as the industrial partner.

UCLA Statistics Club, 2022–2024. Research mentoring for undergraduate and master’s students, helping them formulate tractable questions, learn technical tools, and communicate their results.

For students

If these questions interest you, you are welcome to email me. A brief note about your background, what you would like to explore, and what you hope to learn is a good starting point.

Selected teaching experience

I have taught undergraduate and master’s students across statistics, data science, engineering, and business. I was nominated for UCLA’s Distinguished Teaching Award in 2024.