Scroll

Thinking mathematically Working in AI

Mathematics is how I make sense of things. I want to know which assumptions underlie a conclusion, what happens when they change, and where intuition stops being reliable.

My work increasingly centers on AI. I build and experiment to understand these systems firsthand, and to work out what is worth doing with them.

Selected Research

Critical Dependence Stress for Portfolio Diversification

I develop a variational framework for measuring the dependence stress needed to reverse a portfolio’s Expected Shortfall advantage while preserving individual loss distributions. I prove a quadratic law for the critical stress budget and characterize exactly when marginal restoration adds a premium. In a uniform benchmark, the asymptotic reversal budget is at least 30% higher than the initial-sensitivity prediction. Continuous flow–cut duality characterizes the premium’s sensitivity to background displacement costs, turning local risk sensitivity into a mathematically derived threshold for portfolio decision reversal.

Certified Valuation of Arithmetic Asian Options

I develop a weighted payoff-remainder method that turns an exact conditional price representation into certified arithmetic Asian valuations. Combining shared Gaussian smoothing with validated Heston transforms, I obtain a complete projected Euler discretization-error bound below 0.011025 price units for a one-year call spread with twelve monthly fixings. I also derive a direct coupling bound that reduces the nonlinear remainder width by 99.4% in a deterministic-variance example, with proofs and executable certificates supporting both results.

Certified Joint Pricing Errors in Rough Heston

I develop a deterministic framework for certifying rough Heston portfolio valuations. I derive a finite-history propagation kernel and the exact support function of a joint error enclosure, capturing shared Fourier errors that separate option-by-option bounds lose. In a matched benchmark, finite-history propagation tightens the complete pricing-error upper bound by 72.91% without changing the pricing output. Mathematical proofs and executable certificates connect the theoretical advances to verifiable financial results.

Education

Duke University

Master of Engineering in Risk Engineering

Financial Risk Concentration

Academic Advisor: Mark Borsuk, Ph.D.

Pratt School of Engineering Merit Scholarship

Bachelor of Science in Mathematics

Dual-Degree Undergraduate Program

Undergraduate Merit Scholarship