I teach machines toread the human brain.
I am a postdoctoral research fellow at Penn Medicine working with Prof. Ragini Verma. My research develops AI tools that connect cellular-resolution histopathology with macroscopic brain imaging. I build deep learning methods for segmenting and quantifying axonal injury in APP immunohistochemistry, with the goal of linking microscopic pathology maps to imaging measures of white-matter injury, including MRI, fMRI, diffusion MRI, and tractography-derived measures.
Currently open for collaborations and consulting. say hi →
what i'm working on
High-resolution fetal brain segmentation
Cascade U-Nets that extract the brain from motion-corrupted in-utero MRI. The fetus moves, the brain is small, and it's growing fast — classical methods fail. Synthetic training data fills the gap.
Tractography segmentation for neurosurgical planning
Tractography & white-matter bundling for neurosurgical planning. Knowing exactly which connections to spare in TBI and tumor patients is the difference between recovery and lasting harm.
Connectome alignment with optimal transport
Optimal transport over functional connectomes — for alignment, comparison, and behavior prediction. The math is beautiful and the predictions hold up. This was my PhD thesis work at Yale.
Synthetic medical image generation
Medical datasets are tiny. So we generate them. Gaussian Mixture and Hidden Markov Random Field models produce label maps that synthesize unlimited training images — for any modality you can describe.
Whole-body CT/MRI segmentation
Organ segmentation in CT/MRI — liver, lungs, kidneys. The pipelines I build help radiologists track disease progression, measure organs, and plan surgery across modalities and centers.
Cross-lingual information retrieval
Before medical imaging, I worked on EM-based dictionary translation for cross-lingual retrieval. Resulted in a US patent and publications in COLING and IP&M Journal. Different field, same instincts.