His research focuses on several methodological aspects of simulation modeling and its intersection with optimization and statistics. He is interested in quantifying and mitigating the uncertainty of stochastic performance analysis due to model misspecification, by developing simulation-based machineries for sensitivity and robust analysis. He is also interested in building computation and optimization strategies that blend with input data, their statistical models, and other information such as output validation data, with the goal of offering guarantees in both computational and statistical efficiency. His other lines of work consist of designing techniques to speed up Monte Carlo schemes and simulation optimization algorithms that are applied to risk management, extreme event analysis, and decision-making in service and engineering systems.
This graph shows how the provably best nonparametric worst-case bounds for the overload probability of a queueing system compare to those using various parametric models on the service time distribution. Generating these bounds requires the use of nested simulation and bootstrapping. The bounds can be used as sensitivity tools to measure the risk of adopting misspecified service time model.