people

Hi, welcome! I’m Hyun Seo (Emily), a first-year PhD student in Biomedical Engineering at Johns Hopkins University, advised by Nicholas Durr in the Computational Biophotonics Lab. I work at the intersection of machine learning and medical imaging, with a focus on building diagnostic and sensing systems that are efficient on edge devices, robust across patient populations and imaging conditions, and aware of their own uncertainty.

My current work spans two threads. The first is computational optical imaging: I’m developing a smartphone-based nailfold capillaroscopy system for non-invasive hemoglobin estimation, where sources of variability — skin tone, finger curvature, illumination — are treated as first-class design constraints rather than nuisances to correct for downstream. The second is adaptive inference for medical vision: I co-led a framework for medical Vision Transformers that combines token reduction with early exiting, allocating compute based on per-image difficulty while preserving diagnostic accuracy.

Outside of research, I enjoy playing volleyball and soccer, bingeing Netflix shows, and cooking Korean food. Always happy to chat about computational imaging, robust medical AI, or how research ideas become devices that reach patients — feel free to reach out.

My interests: Medical Imaging, Computational Optical Imaging, Nailfold Capillaroscopy, Vision Transformers, Smartphone Diagnostics, Robust Medical AI, Biomedical Optics, Non-invasive Blood Analysis.

Hi, welcome! I’m Hyun Seo (Emily), a first-year PhD student in Biomedical Engineering at Johns Hopkins University, advised by Nicholas Durr in the Computational Biophotonics Lab. I work at the intersection of machine learning and medical imaging, with a focus on building diagnostic and sensing systems that are efficient on edge devices, robust across patient populations and imaging conditions, and aware of their own uncertainty.

My current work spans two threads. The first is computational optical imaging: I’m developing a smartphone-based nailfold capillaroscopy system for non-invasive hemoglobin estimation, where sources of variability — skin tone, finger curvature, illumination — are treated as first-class design constraints rather than nuisances to correct for downstream. The second is adaptive inference for medical vision: I co-led a framework for medical Vision Transformers that combines token reduction with early exiting, allocating compute based on per-image difficulty while preserving diagnostic accuracy.

Outside of research, I enjoy playing volleyball and soccer, bingeing Netflix shows, and cooking Korean food. Always happy to chat about computational imaging, robust medical AI, or how research ideas become devices that reach patients — feel free to reach out.

My interests: Medical Imaging, Computational Optical Imaging, Nailfold Capillaroscopy, Vision Transformers, Smartphone Diagnostics, Robust Medical AI, Biomedical Optics, Non-invasive Blood Analysis.