Physics undergraduate researcher working at the intersection of quantitative finance, machine learning, and high energy nuclear physics, applying data driven modeling to limit order book dynamics, gluon saturation, geometric scaling, and particle collision analysis with Python, numerical methods, C++/ROOT, and reproducible computational workflows.
His academic work centers on the future Electron-Ion Collider, which will probe the structure of protons and nuclei through the quarks and gluons that make them up. His work helps build expectations for what the EIC should measure and how those measurements may appear.
Uses C++, ROOT, Python, Unix/Linux, numerical simulations, Monte Carlo event generators, and data analysis workflows for physics research.
Interested in paths that apply rigorous analytical methods to complex systems, including graduate research in nuclear physics and quantitative finance. Across both fields, his focus is on computational modeling, data analysis, and careful reasoning under uncertainty.
Projects connecting quantitative finance, machine learning, data analysis, and computational modeling with nuclear physics, gluon saturation, and high-energy collision research.
Completed end-to-end benchmark for classifying short-horizon limit-order-book movement on real FI-2010 data. The selected histogram gradient-boosting model reached 73.25% held-out three-class accuracy, 61.35% balanced accuracy, and 64.13% macro F1—an improvement of 16.31 percentage points and 37.88% lower error than the majority baseline. Chronological model selection, committed diagnostics, and checksummed data retrieval make the result fully reproducible.
Limit Order Books Gradient Boosting FI-2010 Python View results & code →Before researchers can study electron–nucleus collisions, they must reconstruct key kinematic quantities for diffractive procces. This time-intensive step can slow the analysis of large datasets. This project uses machine learning to make reconstruction faster and more accurate, improving data-analysis efficiency across the international electron-Proton/Ion Collider (ePIC) collaboration.
Machine Learning Kinematics ePICAt high energies, gluons multiply until their growth may begin to saturate, revealing new behavior in nuclear matter. This research analyzes 10 million simulated electron–nucleus collisions using geometric scaling and spatially dependent nuclear structure to study where saturation occurs and how the Electron–Ion Collider could detect it.
EIC QCD Gluon Saturation nPDFOngoing EIC study extending an invisible scalar- and vector-boson search with machine learning. The project reproduces generator-level cross-section baselines, builds weighted signal and background samples, and trains boosted decision trees to test whether multivariate correlations improve sensitivity beyond rectangular cuts. The current result is a parton-level proof of concept; detector-level validation is next.
Machine Learning Dark Bosons EIC C++/ROOTFellowships, presentation awards, travel grants, and academic recognition.
Photos from research programs, conferences, presentations, leadership, and academic experiences.
