About
Summary
I am a biomedical engineer and applied ML researcher with a PhD from the University of Wisconsin-Madison. My dissertation work built a closed-loop evaluation framework for generative models trained on second harmonic generation microscopy images of collagen, connecting latent space behavior to quantified morphological features including fiber length, density, and orientation. I am broadly interested in making clinical AI more trustworthy by documenting and measuring failure modes, rather than relying on benchmark accuracy as a proxy for deployment readiness. I am currently seeking postdoctoral and research scientist positions in healthcare-focused ML, imaging, and interpretability.
Positions
PhD Student Aug 2017 - Aug 2025
University of Wisconsin-Madison
Thesis: SHG Imaging and Machine Learning Methods for Feature Extraction and Analysis Tool Development.
- Trained StyleGAN2-ADA-SHG on a curated library of 1,319 SHG microscopy images spanning multiple tissue types and disease states (ovarian cancer, pulmonary fibrosis, connective tissue disorders), adapting the architecture to greyscale input with adaptive discriminator augmentation for effective limited-data training.
- Developed a simulation-augmented benchmarking framework using SeFa semantic factorization and image projection to produce synthetic SHG images with independently controllable morphological properties, then quantitatively verified output against ground truth via CT-FIRE, CurveAlign, and FFT analysis, establishing reliability boundaries for downstream analysis methods [2].
- Built and evaluated two image projection strategies (LPIPS-based optimization and a custom encoder architecture) for mapping real SHG images into the generator’s latent space, characterizing reconstruction fidelity across fiber metrics and identifying where fine-grained biological detail is preserved or degraded.
- Implemented a partial distance correlation pipeline to assess semantic-morphology relationships while controlling for correlated fiber features, and integrated all tools into an extensible Streamlit GUI supporting real-time latent space traversal, semantic weight control, and interactive analysis visualization.
Undergraduate Researcher Feb 2015 - Dec 2016
Developed a freeform subset selection algorithm in MATLAB for deformation mapping techniques and integration with digital image correlation.
Education
University of Wisconsin-Madison 2017 - 2025
Field of study: Biomedical Engineering
Degree: PhD
Oregon State University 2012 - 2016
Field of study: Mechanical Engineering
Degree: Bachelors
Skills
End-to-end pipeline development for biomedical imaging, from data curation and preprocessing through generative model training, latent space analysis, and interactive visualization. Experienced in building rigorous evaluation frameworks under domain-specific constraints, including heterogeneous datasets, limited training data, and grayscale microscopy textures where standard metrics have limited validity. Additional experience in production software engineering including relational schema design, containerized development, and CI/CD workflows.
Keywords: Data Analysis, Docker, git, Image Processing, Linux, Machine Learning, MATLAB, OpenCV, Python, pytorch, Scikit-Professional interests
Applied machine learning for biomedical imaging, with a focus on building systematic evaluation frameworks that connect model behavior to measurable biological structure. Particular interest in generative models, interpretability methods, simulation-augmented benchmarking, and domain shift in clinical imaging contexts. Motivated by problems at the intersection of ML methodology and real clinical consequence, including disease characterization, morphological feature extraction, and multimodal data integration.
Keywords: biomedical image analysis, Cancer Research, disease mechanisms, Feature autodetection, Interpretability, Machine Learning, MultidisciplinaryCV
Please sign in to view the CV.
Contact details
Please sign in to see contact details.New opportunities
Open to new opportunities: