Research

Previous Research Interests

My previous research applied physics-informed machine learning and statistical methods to complex systems, with a focus on:

• Spatio-temporal spectral analysis of geophysical fluid dynamics
• Machine learning emulation of Earth-system models
• High-performance scientific data pipelines for multi-terabyte climate datasets
• Stochastic simulation and statistical inference for biological systems
• Network and graph-theoretic methods for structural characterization of biological materials

As part of the Arbic Lab at the University of Michigan, I developed a spatio-temporal spectral decomposition algorithm for analyzing energy transfers in NASA’s coupled Earth-system simulation (MITgcm/GEOS5). This work involved building parallelized ETL pipelines on NASA Pleiades HPC to process ~1 TB of NetCDF ocean simulation data, and resulted in a publication in Physical Review Fluids (2025).

In the Lubensky Lab at the University of Michigan, I built the lab’s first semi-automated image analysis pipeline (Python, OpenCV, scikit-learn, FIJI) for processing ~1 TB of microscopy data — achieving a 10x throughput improvement. I also engineered a C++ physics-based simulation of 100,000+ heterogeneous cells with stochastic dynamics to study active matter and biophysical systems.

As an undergraduate researcher and Worster Fellow in the Carlson Lab at UC Santa Barbara, I applied graph-theoretic methods to develop novel structural indicators for bone disease diagnostics in human trabecular bone. This work included building semi-automated pipelines to analyze microCT confocal images and generate 3D structural metrics, and resulted in a first-author publication in Physical Review E (2019).