I'm a Postdoctoral Associate at Weill Cornell Medicine, working with Mert Sabuncu and Qingyu Zhao in the Machine Intelligence in NeuroImaging (MINI) lab.
My research focuses on developing machine learning methods for biomedical data, particularly representation learning, self-supervised learning, and multimodal and longitudinal modeling.
I'm interested in developing ML methods that go beyond task-specific predictions to better model biological systems and individual variability, with applications in neuroimaging, medical imaging, and clinical data.
📍 New York City
Google Scholar · LinkedIn · X / Twitter
Continuous representations of brain function (Present)
Exploring implicit neural representations to model fMRI signals and functional connectivity continuously, beyond fixed spatial parcellations and discrete descriptions of brain dynamics.
Longitudinal modeling of multimorbidity trajectories (Present)
Developed transformer-based models of longitudinal electronic health records to forecast comorbidities and investigate patient subtypes and trajectories across substance use, mental health, and chronic physical conditions using All of Us Research Data.
Self-supervised representation learning for functional connectivity (MICCAI 2026)
Developed a transformer-based framework with network-aware bilinear tokenization and masked autoencoding to learn representations of resting-state functional connectivity.
Network-Aware Bilinear Tokenization for Brain Functional Connectivity Representation Learning. Paper · Code
Individual differences in brain function and behavior (2024-Present)
Studied associations between resting-state functional connectivity, behavior, cognition, and psychiatric symptoms across large developmental neuroimaging cohorts.
- Sex-Specific Associations Between Functional Brain Connectivity and Psychopathology — Translational Psychiatry, 2025. Paper · Code
- Socioemotional and Executive Control Mismatch in Adolescence and Risks for Initiating Drinking — JAMA Network Open, 2025. Paper
- Heterogeneous Brain Functional Mapping of Childhood Psychiatric Symptoms — Society for Neuroscience, 2024. Abstract · Selected for Neuroscience Press Conference (top 0.5% of 10,500 abstracts).
My PhD focused on longitudinal 3D MRI for renal transplantation monitoring, combining self-supervised representation learning, multimodal modeling, and clinical prediction.
Radiomics and renal graft survival prediction (Academic Radiology, 2025)
Investigated MRI-derived representations and radiomics for predicting long-term renal graft survival.
Renal transplant survival prediction from unsupervised deep learning-based radiomics on early dynamic contrast-enhanced MRI. Paper
Multimodal MRI representation learning with clinical information (MIDL 2024)
Developed a CLIP-inspired framework using LLM-generated clinical descriptions to align 3D MRI representations with multiple clinical variables and improve renal function forecasting.
MEDIMP: Medical Images and Clinical Prompts for Multimodal Representation Learning. Paper · Code
Contrastive learning and longitudinal MRI modeling (MICCAI 2022)
Developed clinically guided contrastive pretraining of 3D MRI representations, followed by masked transformer modeling to forecast renal function over time.
Contrastive Masked Transformers for Longitudinal 3D Medical Images. Paper · Code
Unsupervised kidney segmentation (ISBI 2021)
Developed unsupervised deep learning methods for kidney segmentation in MRI to support downstream longitudinal imaging analyses.
3D unsupervised kidney graft segmentation based on deep learning and multi-sequence MRI. Paper
LISA Challenge (MICCAI 2026)
Developed a unified deep learning framework for ultra-low-field pediatric brain MRI, combining image enhancement, subcortical structure segmentation, and artifact quality assessment using a multi-task 3D U-Net architecture.
BraTS Challenge (MICCAI 2024)
Contributed to brain tumor segmentation using artificial MRI sequence generation and ensemble learning (Rank 2 on BraTS 2024 Task 1: Segmentation – Adult Glioma Post-Treatment.)
Deep-stULM: Deep Learning for Ultrasound Localization Microscopy (IEEE TMI 2021)
Developed deep learning methods for spatiotemporal super-resolution ultrasound imaging to improve visualization of microvascular structures.
Representation Learning · Self-Supervised Learning · Multimodal Learning · Transformers · Longitudinal Modeling · Neuroimaging · Medical Imaging · AI for Life Sciences
Python · PyTorch · Scikit-learn · Distributed Training (DDP / SLURM) · Medical Image Processing