Susan Wei

Susan Wei

Associate Professor

Department of Econometrics and Business Statistics, Monash University

Biography

Susan Wei is an Associate Professor in the Department of Econometrics and Business Statistics at Monash University. She previously held a Discovery Early Career Researcher Award (DECRA) and was a visiting faculty researcher at Google DeepMind in Sydney. Her research focuses on statistical machine learning, particularly in Bayesian approaches to deep learning, alongside variational inference and singular learning theory.

She is a 2026 recipient of the UK AI Safety Institute Alignment Project grant for her project Probabilistic Interpretability of Transformers.

Interests

  • Statistics
  • Deep Learning
  • Singular learning theory
  • Tabular foundation models

Education

  • PhD in Statistics, 2014

    University of North Carolina, Chapel Hill

  • BA in Mathematics, 2009

    University of California, Berkeley

Grants

Experience

 
 
 
 
 

Associate Professor

Department of Econometrics and Business Statistics, Monash University

Dec 2024 – Present Melbourne, Australia
 
 
 
 
 

Visiting Faculty Researcher

Google Deepmind

Jun 2023 – Dec 2023 Sydney, Australia
 
 
 
 
 

Lecturer (Assistant Professor)

School of Mathematics and Statistics, University of Melbourne

Jun 2018 – Dec 2024 Melbourne, Australia
 
 
 
 
 

Assistant Professor

Division of Biostatistics, University of Minnesota

Jan 2016 – Apr 2018 Minnesota, USA
 
 
 
 
 

Postdoc

Institute of Mathematics, Ecole Polytechnique Federale de Lausanne

Apr 2014 – Dec 2015 Lausanne, Switzerland

Group

Current

  • Paul Roy Lessard — Postdoctoral Researcher, Monash (ongoing)
  • Zhiyuan (Jerry) Xu — PhD, co-supervised with Jack Jewson (2026–)
  • Dilmi Abeytunga — PhD, co-supervised with Russell Tsuchida (2025–)
  • Haotong Ma — PhD (2025–)
  • Kenyon Ng — PhD, primary supervision (2022–2026) → Postdoctoral Researcher, Nanyang Technological University (NTU)
  • Guoyang (Gary) Zheng — Honours, Monash (2026) → Boston Consulting Group X

Alumni and placements

  • Edmund Lau — PhD (2020–2025), co-supervised with Daniel Murfet → Symbolica → UK AI Safety Institute (AISI)
  • Aoqi Zuo — PhD (2021–2025), co-supervised with Mingming Gong → Postdoctoral Researcher, University of Sydney
  • Afiq Aswadi — Master’s (2025) → Research Fellow, Monash → MATS (ML Alignment & Theory Scholars)
  • Hui Li — PhD (2019–2023) → Center for AI in Drug Discovery, Case Western Reserve University

Upcoming Talks

ISBA 2026 World Meeting

Pretrained Transformers as In-Context Bayesian Learners: Implications for Uncertainty Quantification
Jun 28, 2026 Nagoya, Japan

Recent Talks

For a complete list, see my CV.

IMS-APRM 2026

Pretrained Transformers as In-Context Bayesian Learners: Implications for Uncertainty Quantification
Jun 13, 2026 Chinese University of Hong Kong

Bin Yu's group, UC Berkeley

Pretrained Transformers as In-Context Bayesian Learners: Implications for Uncertainty Quantification
Apr 8, 2026 UC Berkeley

Bayes on the Beach 2026

A Principled Framework for Uncertainty Decomposition in TabPFN
Feb 10, 2026 Wollongong

The Second Sydney Workshop on Mathematics of Data Science

Bayesian Predictive Inference and Transformers
Dec 10, 2025 University of Sydney

Melbourne and RIKEN-AIP Joint Workshop

Transformers as Bayesian Predictive Engines
Jul 3, 2025 Melbourne Connect

MATRIX: Bayesian Learning of Very High-Dimensional Physical Process Models

What Is Prediction For? Turning Pre-Trained Transformers into Bayesian Engines
Jun 23, 2025 MATRIX, Creswick

Advances in post-Bayesian methods, UCL

What Is Prediction For? Turning Pre-Trained Transformers into Bayesian Engines
May 1, 2025 UCL

Object Oriented Data Analysis (Marron 70th Birthday)

Dynamical versus Bayesian Transitions in a Toy Model of Superposition
May 1, 2025 UNC Chapel Hill

ETH Zurich Seminar

What’s Degeneracy Got to Do with It? Understanding Deep Neural Networks through Singular Learning Theory
Apr 1, 2025 ETH Zurich

Mostly Monte Carlo Seminar Series, PariSante

Dynamical versus Bayesian Transitions in a Toy Model of Superposition
Apr 1, 2025 PariSante, Paris
See all talks

Selected Workshop Invitations

Workshop Organisation

Teaching

Deep learning

I have developed and taught an introduction to deep learning across several rounds of the Australian Mathematical Sciences Institute (AMSI) national schools, which are open to graduate students, early career researchers, and industry members across Australia:

  • AMSI Summer School 2024 — a new course on the theoretical foundations of deep learning (co-taught with Pavel Krupskiy and Matthew Tam).
  • AMSI Winter School 2021 — Neural Networks and Related Models, covering the deep learning pipeline alongside probabilistic models involving neural networks.

My lecture slides for the introductory deep learning material are available here, covering:

  • An introduction to neural networks: key components of the DL pipeline, multilayer perceptrons, forward/backward propagation, computational graphs
  • Stochastic optimization and extensions
  • The art of model training and regularization: model selection, weight decay, dropout, initialization
  • Convolutional and recurrent neural networks

The companion module on deep generative modeling was given by Robert Salomone, whose materials are here.

University coursework

Selected units I have taught (see the linked handbook entries for details):

Monash University

University of Melbourne