Omar Bennouna

Omar Bennouna is a graduate student in computer science establishing principled frameworks for determining which data actually matter for decision-making problems. Organizations collect vastly more data than necessary; his core question: given a linear program with uncertain cost parameters, which observations suffice to recover the optimal decision? He provides a complete geometric characterization showing that data sufficiency depends on whether the data span covers all cost directions that could alter the optimal solution. Using MATLAB for algorithm prototyping, numerical experiments, and visualization, he developed tractable algorithms for constructing minimal sufficient datasets. Applications include strategic field-study selection for infrastructure design and non-adaptive interview scheduling for optimal hiring. As a MathWorks Fellow, Omar will extend this framework beyond linear programs to mixed-integer and convex programs with stochastic observations. His work could fundamentally reshape data collection practices by establishing task-aware acquisition methods grounded in decision geometry rather than convenience.