In engineering, surrogate models offer a fast and efficient alternative to conventional high-fidelity simulations. They enable early concept validation and interactive design through real-time predictions. The development process typically involves data sampling, model training, and validation—often based on results from computationally expensive simulations. Depending on the application, different approaches such as physics-based, data-driven, or hybrid models are used, each requiring distinct training strategies. Surrogate models are widely applicable across domains such as finite element analysis (FEM), multibody simulation (MBS), and computational fluid dynamics (CFD), and can predict scalar values, time series, or even mesh deformations.