This work introduces the DiMAT simulation toolkits, a set of Multiphysics and multiscale solutions designed to predict material behaviour and optimize manufacturing processes in real industrial contexts. By combining physics-based models (including LAMMPS for atomistic simulations and CalculiX for finite element analysis) with artificial intelligence, the toolkits form a hybrid modelling framework capable of accurately capturing complex phenomena across multiple scales, from the molecular level through component-scale performance. A core feature of these toolkits is their ability to integrate process simulation (e.g., polymer extrusion, melt spinning, yarn and net formation) with advanced material models, facilitating a seamless transition from raw composition data to final product performance assessment. The multi-fidelity approach supports real-time computation and enables rapid design iterations without compromising on accuracy. The resulting predictive capabilities are further enhanced through machine learning techniques, allowing for continuous improvement and near-instant feedback when operating in a Digital Twin environment. Industrial adopters benefit from streamlined workflows, since each toolkit can be accessed via intuitive interfaces and easily customized for specific pilot cases. Moreover, the modular design supports flexible integration reducing time-to-deployment. Ultimately, the DiMAT solution empowers engineers and researchers to gain deeper insights into material behaviour, shorten development cycles, and refine manufacturing processes to achieve superior product performance at scale, all while preserving a robust connection to the underlying physical reality.