Data-inspired and hybrid physics-data techniques are being applied across a wide spectrum of nonlinear systems, enhancing capabilities in modeling, simulation, prediction, and optimization. These methods provide powerful new ways to uncover hidden patterns, develop predictive models, and manage the inherent complexities of dynamical systems. Strategies exploiting neural networks, deep learning, and hybrid physics-data architectures (e.g., physics-inspired symbolic regression, deep symbolic regression methods) which merge physical insights with machine learning to derive interpretable models are particularly welcome. The colloquium will provide opportunities for discussion, collaboration, and exploration of new research directions in machine learning for nonlinear dynamics.