This chapter gives an overview of the applications of machine learning in computational mechanics, focusing on surrogate models that accelerate complex simulations while maintaining accuracy. Machine learning approaches—including feedforward neural networks (FNN), recurrent neural networks (RNN), convolutional neural networks (CNN), and generative adversarial networks (GAN)—are systematically examined across several domains. The chapter covers multiscale and constitutive modeling, where neural networks replace expensive representative volume element computations; structural topology optimization, where deep learning accelerates design iterations; partial differential equation solvers enhanced by physics-informed neural networks; improvements to finite element method solutions; and computational fluid dynamics acceleration through various architectures. Special attention is given to reduced-order modeling techniques that combine traditional dimensionality reduction methods like proper orthogonal decomposition with advanced neural network architectures, achieving computational speedups of several orders of magnitude. The chapter concludes with a discussion of digital twins—virtual representations of physical systems through continuous data exchange. Applications span predictive maintenance, structural health monitoring, and cyber-physical systems, demonstrating how machine learning-enhanced simulations enable practical engineering solutions previously constrained by computational costs.