Due to their spatial and temporal variability, submesoscales occupy a unique dynamical regime, bridging large-scale quasi-2D turbulence and small-scale 3D turbulence. These properties pose significant challenges for theoretical, numerical, and observational efforts to study submescale processes. In this talk, I will present insights from idealized and global simulations, the integration of observational data, and the application of machine learning—each aimed at improving our understanding of submesoscale turbulence and its interactions with mesoscale and larger-scale ocean dynamics, as well as upper-ocean boundary layer turbulence.