Turbulence plays important roles in mission-critical systems such as aircraft and jet engines, but the prediction of turbulence continues to defy and bedevil scientists and engineers. Recent advances in flow diagnosis and high-fidelity simulations generated enormous amounts of data, opening up new opportunities. Our research leverages data science (uncertainty quantification, data assimilation, machine learning) to advance understanding and improve predictions of turbulent flows and other multi-scale physical systems.