Probabilistic Modeling and Scientific Machine Learning for Computational Physics and Engineering

This PhD thesis integrates probabilistic machine learning with physics-based modeling to improve computational methods in science and engineering. It presents strategies to address challenges such as limited data, noise, and differentiability. Key contributions include a probabilistic data-driven turbulence model, a framework for sustainable concrete design combining model discovery and optimization, and a novel algorithm for optimizing high-dimensional, physics-based simulations.