Dongwon Han

Applied AI/ML — Machine Learning Engineering

J.P. Morgan

London, United Kingdom

dwhan89[at]gmail.com

I work in Applied AI/ML at J.P. Morgan, building and deploying robust, scalable machine learning systems. I am PhD-trained in computational physics, with a background in probabilistic modelling, simulation-based inference, uncertainty estimation, and high-performance computing.

My interests span simulation-driven, probabilistic, and multimodal machine learning, particularly for complex decision-making under uncertainty. Before moving into industry, I worked on cosmic microwave background (CMB) data analysis as a member of the Atacama Cosmology Telescope (ACT) and Simons Observatory (SO) collaborations, and as a postdoctoral researcher at the University of Cambridge (DAMTP) — see my experience and publications for more.

news

Aug 1, 2026	Started a new role in Applied AI/ML — Machine Learning Engineering at J.P. Morgan in London.
Dec 1, 2021	a beginning of new blog

selected publications

JCRR
Assessing Climate Change Impacts in Hurricane Catastrophe Models Using a Frequency Aware Quantile Mapping

Petrie, Ruth, Han, Dongwon, Nicotina, Ludovico, Alvarez, Adam, and Cox, Tyler

Journal of Catastrophe Risk and Resilience 2026

Abs Bib

We present a frequency-aware quantile mapping (FAQM) approach for adjusting hurricane wind speed distributions in catastrophe models under climate change scenarios. Traditional approaches typically resample events to increase the frequency of high-category hurricanes; FAQM instead preserves the original event set structure while shifting severity distributions toward climate-conditioned targets, offering a practical alternative when the number of events in an event set must remain fixed.
@article{Petrie2026, bibtex_show = {true}, selected = {true}, author = {Petrie, Ruth and Han, Dongwon and Nicotina, Ludovico and Alvarez, Adam and Cox, Tyler}, doi = {10.63024/rf3b-sgcj}, journal = {Journal of Catastrophe Risk and Resilience}, abbr = {JCRR}, title = {Assessing Climate Change Impacts in Hurricane Catastrophe Models Using a Frequency Aware Quantile Mapping}, volume = {04}, pages = {02}, url = {https://journalofcrr.com/research/04-02-petrie-et-al/}, year = {2026} }
arxiv
Recovering Galaxy Cluster Convergence from Lensed CMB with Generative Adversarial Networks

Parker, Liam, Han, Dongwon, Portela, Pablo Lemos, and Ho, Shirley

arXiv e-prints Nov 2022

Abs Bib

We present a machine learning approach using conditional Generative Adversarial Networks (cGANs) to reconstruct galaxy cluster convergence maps from lensed cosmic microwave background (CMB) temperature maps. Relative to a Residual U-Net baseline, the cGAN model emphasizes structure and high-frequency correctness, demonstrating improved performance when processing noisy CMB data, particularly at high frequencies where traditional methods have struggled.
@article{Parker2022, bibtex_show = {true}, selected = {true}, author = {Parker, Liam and Han, Dongwon and Portela, Pablo Lemos and Ho, Shirley}, month = nov, journal = {arXiv e-prints}, abbr = {arxiv}, title = {Recovering Galaxy Cluster Convergence from Lensed CMB with Generative Adversarial Networks}, url = {https://arxiv.org/abs/2211.08990}, year = {2022} }

PRD

Deep learning simulations of the microwave sky

Han, Dongwon, Sehgal, Neelima, and Villaescusa-Navarro, Francisco

Physical Review D Dec 2021

Bib

@article{Han2021b,
  bibtex_show = {true},
  selected = {true},
  author = {Han, Dongwon and Sehgal, Neelima and Villaescusa-Navarro, Francisco},
  doi = {10.1103/PhysRevD.104.123521},
  issue = {12},
  journal = {Physical Review D},
  abbr = {PRD},
  month = dec,
  pages = {123521},
  publisher = {American Physical Society},
  title = {Deep learning simulations of the microwave sky},
  volume = {104},
  url = {https://journals.aps.org/prd/abstract/10.1103/PhysRevD.104.123521},
  year = {2021}
}

arxiv
Mitigating Foreground Bias to the CMB Lensing Power Spectrum for a CMB-HD Survey

Han, Dongwon, and Sehgal, Neelima

arXiv e-prints Dec 2021

Abs Bib

A promising way to measure the distribution of matter on small scales (k 10 hMpc^-1) is to use gravitational lensing of the Cosmic Microwave Background (CMB). CMB-HD, a proposed high-resolution, low-noise millimeter survey over half the sky, can measure the CMB lensing auto spectrum on such small scales enabling measurements that can distinguish between a cold dark matter (CDM) model and alternative models designed to solve problems with CDM on small scales. However, extragalactic foregrounds can bias the CMB lensing auto spectrum if left untreated. We present a foreground mitigation strategy that provides a path to reduce the bias from two of the most dominant foregrounds, the thermal Sunyaev-Zel’dovich effect (tSZ) and the Cosmic Infrared Background (CIB). Given the level of realism included in our analysis, we find that the tSZ alone and the CIB alone bias the lensing auto spectrum by 0.6 sigma and 1.1 sigma respectively, in the lensing multipole range of L in [5000,20000] for a CMB-HD survey; combined these foregrounds yield a bias of only 1.3 sigma. Including these foregrounds, we also find that a CMB-HD survey can distinguish between a CDM model and a 10^-22 eV FDM model at the 5 sigma level. These results provide an important step in demonstrating that foreground contamination can be sufficiently reduced to enable a robust measurement of the small-scale matter power spectrum with CMB-HD.
@article{Han2021c, bibtex_show = {true}, selected = {true}, author = {Han, Dongwon and Sehgal, Neelima}, month = dec, journal = {arXiv e-prints}, abbr = {arxiv}, title = {Mitigating Foreground Bias to the CMB Lensing Power Spectrum for a CMB-HD Survey}, url = {https://arxiv.org/abs/2112.02109v1}, year = {2021} }
JCAP
The Atacama Cosmology Telescope: delensed power spectra and parameters

Han, Dongwon, Sehgal, Neelima, MacInnis, Amanda, Engelen, Alexander, Sherwin, Blake D., Madhavacheril, Mathew S., Aiola, Simone, Battaglia, Nicholas, Beall, James A., Becker, Daniel T., Calabrese, Erminia, Choi, Steve K., Darwish, Omar, Denison, Edward V., Devlin, Mark J., Dunkley, Jo, Ferraro, Simone, Fox, Anna E., Hasselfield, Matthew, Hill, J. Colin, Hilton, Gene C., Hilton, Matt, Hložek, Renée, Hubmayr, Johannes, Hughes, John P., Kosowsky, Arthur, Lanen, Jeff Van, Louis, Thibaut, Moodley, Kavilan, Naess, Sigurd, Namikawa, Toshiya, Nati, Federico, Nibarger, John P., Niemack, Michael D., Page, Lyman A., Partridge, Bruce, Qu, Frank J., Schillaci, Alessandro, Spergel, David N., Staggs, Suzanne, Storer, Emilie, and Wollack, Edward J.

Journal of Cosmology and Astroparticle Physics Jan 2021

Bib
@article{Han2021a, bibtex_show = {true}, selected = {true}, author = {Han, Dongwon and Sehgal, Neelima and MacInnis, Amanda and van Engelen, Alexander and Sherwin, Blake D. and Madhavacheril, Mathew S. and Aiola, Simone and Battaglia, Nicholas and Beall, James A. and Becker, Daniel T. and Calabrese, Erminia and Choi, Steve K. and Darwish, Omar and Denison, Edward V. and Devlin, Mark J. and Dunkley, Jo and Ferraro, Simone and Fox, Anna E. and Hasselfield, Matthew and Hill, J. Colin and Hilton, Gene C. and Hilton, Matt and Hložek, Renée and Hubmayr, Johannes and Hughes, John P. and Kosowsky, Arthur and Lanen, Jeff Van and Louis, Thibaut and Moodley, Kavilan and Naess, Sigurd and Namikawa, Toshiya and Nati, Federico and Nibarger, John P. and Niemack, Michael D. and Page, Lyman A. and Partridge, Bruce and Qu, Frank J. and Schillaci, Alessandro and Spergel, David N. and Staggs, Suzanne and Storer, Emilie and Wollack, Edward J.}, doi = {10.1088/1475-7516/2021/01/031}, issn = {1475-7516}, issue = {01}, journal = {Journal of Cosmology and Astroparticle Physics}, abbr = {JCAP}, keywords = {Astrophysics - Cosmology and Nongalactic Astrophysics,High Energy Physics - Phenomenology}, month = jan, pages = {031-031}, title = {The Atacama Cosmology Telescope: delensed power spectra and parameters}, volume = {2021}, url = {https://iopscience.iop.org/article/10.1088/1475-7516/2021/01/031}, year = {2021} }