|
Donghyun Kim I graduated this spring from the Georgia Institute of Technology with a B.S. in Computer Science (Machine Learning & Theory Specializations). Broadly, I am interested in machine learning for biomedicine. I have recently joined the Lindert lab at UCLA studying protein structure prediction and de novo protein design as a Rosetta Commons RaMP Post-baccalaureate Research Fellow. I am fortunate to have been advised by Dr. Lisa Brown, Dr. Uygar Sümbül, Dr. Vince Calhoun, and Dr. Srijan Kumar. |
|
Research |
|
Domain Adaptation for 3D Cell Segmentation
Donghyun Kim, Hayden Nunley Blog Post and Poster, 2025 3D cell segmentation is vital for studying embryonic development, but models struggle as ground truth annotations are difficult to obtain. We explore various domain adaptation methods to solve this issue. |
|
Auxiliary objectives improve generalization performance but reduce model specification for low-data neuroimaging-based brain age prediction
Donghyun Kim, Eloy Geenjaar, Vince Calhoun NeurIPS Workshop (UniReps), 2024 We add an auxiliary objective to a brain age prediction model that significantly improves model performance and generalization under low-data regimes. |
|
Methods for analysis of spatial peptidergic communication networks
Donghyun Kim, Rohan Gala Blog Post and Poster, 2024 Peptidergic communication networks in spatial transcriptomic data can be viewed as directed multilayer graphs. We develop computational methods to study the organization of these graphs. |
|
Providing context: Extracting non-linear and dynamic temporal motifs from brain activity
Eloy Geenjaar, Donghyun Kim, Vince Calhoun PLOS One, 2024 bibtex We introduce a temporal representation learning model for resting-state fMRI data that can find novel and interesting temporal motifs that relate brain activity to schizophrenia diagnosis. |
|
UNIGUARD: Towards Universal Safety Guardrails for Jailbreak Attacks on Multimodal Large Language Models
Sejoon Oh, Yiqiao Jin, Megha Sharma, Donghyun Kim, Eric Ma, Gaurav Verma, Srijan Kumar Deployable AI workshop at AAAI 2025 Reliable and Responsible Foundation Models workshop at ICML 2025 bibtex We propose UNIGUARD, a novel multimodal safety guardrail that jointly considers the unimodal and cross-modal harmful signals. |
|
Uncovering the latent dynamics of whole-brain fMRI tasks with a sequential variational autoencoder
Eloy Geenjaar, Donghyun Kim, Riyasat Ohib, Marlena Duda, Amrit Kashyap, Sergey Plis, Vince Calhoun NeurIPS Workshops (DGM4H, Medical Imaging Meets NeurIPS), 2023 bibtex We uncovered dynamical system representations of whole-brain activity with a sequential VAE that improve behavioral predictions. |
|
Recommender System Attribution
Sejoon Oh, Donghyun Kim, Srijan Kumar Blog Post, 2022 - 2023 We explained and improved the performance of recommender systems with model interpretability methods. |
Research ExperienceRosetta Commons RaMP Post-baccalaureate Research Fellow (Lindert Lab @ UCLA) Post-baccalaureate Research Fellow Aug. 2025 - Present Protein structure prediction models Flatiron Institute, Simons Foundation (Center for Computational Biology) Research Intern May. 2025 - Aug. 2025 Diffusion models, Generative adversarial networks, 3D cell segmentation models Allen Institute (Brain Science) Research Intern May. 2024 - Present Graph neural networks, Graph embedding methods, Feature selection, Spatial transcriptomics TReNDS Center (Georgia Tech, Emory, and GSU) Research Assistant Aug. 2022 - May 2025 Variational autoencoders, 3D Convolutional Neural Networks, Underspecification, Neuroimaging CLAWS Lab (Georgia Tech) Research Assistant Aug. 2021 - May 2024 Large transformer models, Vision-language models, Model interpretability, Recommender systems |
Industry ExperienceToast Software Engineer Intern, Payments - Data Team May 2023 - Aug. 2023 Contributed to Apache Camel applications NCR Software Engineer Intern, Computer Vision and Machine Learning Team May 2022 - Aug. 2022 Spearheaded development of new ML models, implementing a Vision Transformer model to classify scanned images of produce The Home Depot Research Intern, Data Science Aug. 2021 - May 2022 Implemented/evaluated various transformer encoders to generate session representations for recommendation systems |
Honors and AwardsGeorgia Tech’s President’s Undergraduate Research Award Recipient 2024 Korean-American Scientists and Engineers Association Undergraduate Scholarship Recipient 2024 IEEE Healthcare Summit Data Hackathon (scRNA-seq track) Runner-up 2021 |
Leadership and VolunteeringPresident | The Artificial Intelligence for Medicine and Healthcare Society 2022-2023 Vice President of Finance | The Artificial Intelligence for Medicine and Healthcare Society 2021-2022 Undergraduate Research Ambassador | Undergraduate Research Opportunities Program 2022-2023 Vice President of Outreach | Korean-American Scientists and Engineers Association 2024-2025 |