Published Research & Preprints
NMF Regularization Techniques for Unmixing Frequency Comb Data
Brendan B. Connelly, Adrian Derderian, Nafiz Faiaz, Henry Tang, Matthew J. Tyler, Jack Diab, Prineha Narang, and Andrea L. Bertozzi
Proc. SPIE 13920, Quantum Sensing, Imaging, and Precision Metrology IV, 139200Y (March 5, 2026)
View Paper on SPIE Digital Library
DOI: 10.1117/12.3081978
Unimodality of q-Fibonomial Coefficients for Small Cases
Brendan B. Connelly, Ezekiel Ito, Thomas C. Martinez, Olha Shevchenko, and Kacey Yang
arXiv:2605.12822 [math.CO] (May 2026)
Proves unimodality of q-Fibonomial coefficients for n ≤ 3, partly resolving a conjecture of Bergeron–Ceballos–Küstner via combinatorial and algebraic methods.
DOI: 10.48550/arXiv.2605.12822
Cross-Asset Order Flow Imbalance Networks for Return Prediction
Graduate course project, MATH 279: Data Science & Machine Learning for Finance
Supervised by Prof. Mihai Cucuringu (Spring 2026)
Modeled cross-asset return prediction using structured OFI-based cross-impact matrices under ridge and lasso regularization; methods include PCA compression and spectral denoising.
Modular Agent with Reflective Search for Systematic Trading
Graduate course project, MATH 285J: Agentic AI for Autonomous Research in Quantitative Finance
Supervised by Prof. Mihai Cucuringu (Spring 2026)
Built a modular LLM/Monte Carlo tree search framework for systematic strategy discovery, testing whether a Deflated Sharpe Ratio–based robustness reward with regime-stability and capacity-cost penalties selects strategies that generalize better than raw in-sample Sharpe; methods include UCT tree search, a reflective LLM agent pipeline, and out-of-sample walk-forward evaluation.