cv

Curriculum vitae of Hyun Seo (Emily) Lee — PhD student in Biomedical Engineering at Johns Hopkins University, Computational Biophotonics Lab.

General Information

Full Name Hyun Seo (Emily) Lee
Email hlee267@jh.edu
Affiliation Johns Hopkins University, Department of Biomedical Engineering
Advisor Nicholas Durr, Ph.D. (Computational Biophotonics Lab)
Languages English, Korean

Education

  • August 2026 - Present
    Ph.D. in Biomedical Engineering
    Johns Hopkins University, Baltimore, MD
    • {"Advisor"=>"Nicholas Durr, Ph.D., Computational Biophotonics Lab."}
    • Research focus: smartphone-based nailfold capillaroscopy and Monte Carlo light-transport modeling for non-invasive hemoglobin estimation.
  • May 2026
    M.S.E. in Biomedical Engineering
    Johns Hopkins University, Baltimore, MD
  • August 2021 – May 2025
    B.S. in Biomedical Engineering and Computer Science
    Johns Hopkins University, Baltimore, MD
    • GPA 3.90 / 4.00. Dean's List (all eligible semesters), Tau Beta Pi, Alpha Eta Mu Beta.
    • Coursework: Biomedical Data Science, Computer Integrated Surgery, Machine Learning for Medical Applications, Deep Learning, Computer Vision, Data Structures, Artificial Intelligence.

Selected Publications

  • 2026
    • V. W. Wong, E. C. Kelly, E. Leow, I.-A. Kroumov, N. Yan, B. Miller, **H. S. Lee**, S. Gottipati, R. Venkatesh, A. L. Robin, S. Kavitha, T. V. Johnson, S. Upadhyaya, K. Parikh. *Patient Perspectives on Home Tonometry Among United States and Indian Populations.* [Investigative Ophthalmology & Visual Science 67, 2316 (ARVO 2026)](https://iovs.arvojournals.org/article.aspx?articleid=2813719).
    • J. Byun*, **H. Lee***, J. Shuff, R. Venkatesh, N. Shekhawat, K. Parikh, R. Chellappa. *Adaptive Inference for Medical Vision Transformers: When to Reduce Tokens and When to Exit Early.* **Medical Imaging with Deep Learning (MIDL) 2026.** (*equal contribution)

Research Experience

  • March 2023 – Present
    Graduate Researcher, Computational Biophotonics Lab
    Johns Hopkins University · Advisor, Nicholas Durr, Ph.D.
    • Built a smartphone-based nailfold capillaroscopy system using green-LED back-illumination and a reverse-lens optical relay; characterized at 25 lp/mm resolution with field-of-view sufficient to resolve 20 μm capillary loops.
    • Worked onsite with Samsung Mobile imaging engineers to characterize the production smartphone camera stack and proposed an exposure-controlled, auto-focus-swept acquisition pipeline that satisfied Samsung's stock-camera constraints while preserving capillary contrast across skin tones.
    • Designed and 3D-printed an ergonomic finger fixture that reduced motion-artifact rejection rates and improved frame-to-frame registration in longitudinal recordings.
    • Implemented an automated focus-sweep routine (focus value 0.0–1.0) to capture capillary loops at multiple focal planes within a single acquisition.
    • Built a Python pipeline (NumPy, OpenCV) for Beer–Lambert hemoglobin estimation from capillary image stacks, including background flat-field correction and vessel segmentation.
    • Ran a pilot study across light- and dark-skinned subjects, identified an exposure-compensation regime that preserves capillary contrast in darker skin, and defined a skin-tone–aware acquisition protocol now used as the lab default.
    • Currently extending the pipeline with Monte Carlo (mcxyz) forward models of light transport in skin to build calibration tables that map observed contrast and apparent vessel width to capillary depth and hemoglobin concentration.
  • May 2025 – Present
    Graduate Researcher, Artificial Intelligence for Engineering and Medicine Lab
    Johns Hopkins University · Advisor, Rama Chellappa, Ph.D.
    • Led design of a unified adaptive-inference framework for DeiT-S Vision Transformers (PyTorch, timm) that combines token reduction and early exit for medical image classification.
    • Integrated four token-reduction strategies (EViT, Top-K, Random, ToMe) and trained a lightweight CNN redundancy predictor with dataset-specific confidence thresholds to decide per-image whether to prune tokens or exit early.
    • Built training, validation, and profiling pipelines across five medical-image datasets, running all FLOPs/latency analyses, ablation studies, and redundancy-vs-accuracy trade-off curves.
    • Co-first-author paper accepted to MIDL 2025.

Industry Experience

  • March 2023 – Present
    Co-founder
    OcuSound · Advisors, Nicholas Durr, Ph.D. and Kunal Parikh, Ph.D.
    • Designed a low-cost acoustic tonometer for at-home intraocular-pressure (IOP) monitoring, targeting longitudinal IOP tracking for glaucoma management.
    • Drove BOM cost below $100 per unit through circuit redesign and component substitution while maintaining clinically relevant IOP correlation in benchtop validation.
    • Wrote the Python signal-processing pipeline that extracts IOP-correlated features from synchronized infrared-sensor and microphone signals.
    • Led an IRB-approved feasibility study at a partner clinic in India, coordinating consent, survey administration, and clinician engagement to evaluate device usability and acceptance.
    • Integrated an SMS-based adherence reminder system to improve patient follow-through in resource-limited settings.
    • Raised $25,000 in non-dilutive funding through translational-research grants and pitch competitions, including a 1st-place finish in the Hopkins New Venture Challenge.
  • June 2024 – August 2024
    R&D New Product Development Intern
    STERIS Endoscopy, Conroe, TX · Advisor, Heba Hijazi
    • Redesigned custom test fixtures in SolidWorks, cutting fixture preparation time by 40% and standardizing setup across the new-product-development team.
    • Built an automated test rig using solenoid valves and relays that reduced manual data-tracking effort by 80% during reliability testing.
    • Presented fixture and sensor-development updates to VP-level R&D leadership.
  • May 2023 – May 2024
    Engineering Intern
    Clear Guide Medical, Baltimore, MD · Advisor, Alican Demir, Ph.D.
    • Automated needle-segmentation in ultrasound frames using Python and a CNN classifier to support more accurate needle-track overlay during interventional procedures.
    • Built a LaTeX + Python documentation toolchain that cut audit-prep time by 30%.
    • Validated 2,000+ device samples with shape-detection software to confirm registration accuracy.

Projects

  • February 2025 – May 2025
    Deep Learning for Melanoma Diagnosis on Dark-Skinned Patients
    • Built a VGG-19 style-transfer pipeline that synthesized darker-skinned augmentations of the ISIC dermoscopy dataset to address skin-tone class imbalance in melanoma classification.
    • Trained a baseline + augmented classifier and compared per-Fitzpatrick-tone accuracy, showing a measurable accuracy gain on the underrepresented dark-skin subset.
    • Implemented a Monk Skin Tone–based filter using Otsu thresholding to mask lesion artifacts during preprocessing.
  • March 2022 – May 2023
    Sleep Apnea Prevention Device
    • Co-designed an undergraduate capstone device that delivers a controlled jaw-thrust within 60 s for patients with sleep-apnea-related breathing cessation.
    • Conducted 25 user interviews and a literature review of jaw-positioning interventions to drive design requirements; iterated three prototype generations through user testing.

Honors and Awards

  • 2025
    • KASF Scholarship Award ($3,000)
  • 2024
    • Innovation & Entrepreneurship Challenge Award ($10,000)
    • Hopkins New Venture Challenge, 1st Place ($5,000)
    • VentureWell E-Team Program ($5,000)
    • Undergraduate Catalyst Award ($1,000)
    • NIH DEBUT Challenge, Honorable Mention ($1,000)
  • 2024 - 2025
    • Tau Beta Pi Engineering Honor Society, top 1/5 of Whiting School by GPA
    • Alpha Eta Mu Beta National Biomedical Engineering Honor Society, top 1/3 of BME by GPA

Leadership and Service

  • March 2023 – Present
    Treasurer, Marketing Team
    The MedTech Network, Johns Hopkins University
    • Managed annual budget for a 50-member organization and streamlined purchase-order workflow for executive events.
    • Promoted 6 networking events that engaged 150+ students and generated 200+ industry contacts.
  • June 2022 – August 2022
    Promotional Team Co-Lead
    MedHacks, Johns Hopkins University
    • Ran social-media outreach for a hackathon that hosted 200+ student participants.
    • Designed event graphics in Photoshop and Adobe Fresco.

Skills

  • Programming
    • Python (NumPy, Pandas, PyTorch, scikit-learn, OpenCV, timm)
    • C, C++, Java, MATLAB, Assembly, LaTeX
  • Machine Learning and Computer Vision
    • Vision Transformers, CNNs, token reduction, early-exit inference
    • Self-supervised learning, transfer learning, dataset-bias analysis
    • Monte Carlo light-transport modeling (mcxyz, MCML)
  • Hardware and Prototyping
    • SolidWorks, Onshape, circuit design, soldering, 3D printing, laser cutting
    • Raspberry Pi, Arduino, smartphone-based optical bench design
  • Imaging and Sensing
    • Optical design, infrared and acoustic sensing, biomedical device prototyping
    • Smartphone camera characterization, exposure and focus control