Sanaa Mouzahir

Sanaa Mouzahir is a graduate student in computational science and engineering and mechanical engineering combining deep learning with rigorous probabilistic methods to improve coarse-grid simulations of fluid flows. Standard closure models are deterministic and provide no uncertainty estimates, limiting their reliability for mission-critical predictions. Sanaa generates training data using the MSEAS finite volume solver, implements stochastic boundary conditions, and runs simulations in parallel. Her Deep Gaussian closure achieves calibrated uncertainty while gracefully degrading on out-of-distribution inputs. She processes hundreds of simulations efficiently using parallel workflows, reducing typical computation times by 60%. The framework integrates numerical solvers with Python learning systems, preserving numerical fidelity during long simulation rollouts. As a MathWorks Fellow, Sanaa will design multi-objective optimization frameworks to balance accuracy against computational cost and develop closure models of transient fluid dynamical systems. Her work opens new possibilities for reliable uncertainty quantification across ocean and environmental modeling applications.