Simon D.A. Thomas
Papers
https://orcid.org/0000-0001-7911-1659https://scholar.google.com/citations?user=kKonjBYAAAAJ
Web of Science ResearcherID: ABE-8823-2021
Letters collected so far
BA MSci MRes MA (Cantab), PhD (Cantab, minor corrections)
Summary
I recently completed my PhD at the University of Cambridge at the Department of Applied Mathematics and Theoretical Physics, funded through the @ai4er-cdt MRes+PhD program. I passed my viva in January 2026, subject to minor corrections.
My thesis The Potential Height of Tropical Cyclone Storm Surges asked what the worst possible hurricane storm surge given a climate is, and how this could be useful for risk assessment and for testing the extreme extrapolation of deep learning surrogates.
I studied Natural Sciences (Physics, BA+MSci) at the University of Cambridge. I am currently an ML researcher at Goldman Sachs.
Projects that I am currently involved in
“Finding the potential height of tropical cyclone storm surges in a changing climate using Bayesian optimization” Read our preprint https://doi.org/10.31223/X57T5R and code https://github.com/sdat2/worstsurge (docs) to hear about a method for calculating the worst possible storm surge and its benefits!
SurgeNet: A spatio-temporal graph neural network for emulating storm surge models. Datasets available on HuggingFace: train/val/test, extreme test.
Feel free to get in touch. My email address is ${my github handle} at cam dot ac dot uk .
Funding
My PhD was supported by studentship 2413578 from the UKRI Centre for Doctoral Training in Application of Artificial Intelligence to the study of Environmental Risks (grant no. EP/S022961/1). I also received funding from the NERC ACSIS project (grant no. NE/N018028/1), and a Natural Environment Research Council (NERC) Research Experience Placement (REP) project funded by the SPITFIRE Doctoral Training Partnership (grant no. NE/S007210/1).
Reviews
I have reviewed articles for:
- Journal of Physical Oceanography (JPO) (x1)
- Journal of Advances in Modeling Earth Systems (JAMES) (x1)
I have reviewed abstracts for:
- Climate Informatics 2024 (x2)
Blog
- London has better weather than New York City — a data-driven analysis of 20 years of hourly weather records that debunks the “drizzly London” myth.
Past projects
- Goldman Sachs: ML research in Model Risk Management — probabilistic model benchmarking, proper scoring rules, config-driven pipelines (Hydra/MLflow).
- Risk Management Solutions (UK) Ltd.: Deep learning for extreme wind superresolution/downscaling.
- Detecting fronts in the Southern Ocean using different algorithms.
- Detecting habitat fragmentation using ML algorithms on earth observation products.
- Using sensitivity analysis on simplified global physical models to understand CMIP biases in the tropics.
- Looking at the return periods of storm surges produced by tropical cyclones in climate models.