Aircraft and helicopter turbine blades operate under extreme thermomechanical loads, where their microstructure critically influences plastic strain localization, creep resistance, oxidation resistance, and fatigue strength. Typically composed of Ni-based superalloys such as IN718, their mechanical response is often assessed through extensive experimental campaigns or numerical homogenization. However, these approaches may fail to fully capture the complexity of in-service behavior, necessitating full-scale simulations that explicitly account for microstructural effects. This work presents a computational framework for full-scale turbine blade simulations that directly incorporate microstructural details. Several key challenges are addressed, including realistic microstructure generation, advanced crystal plasticity modeling with grain size effects, and the solution of large, highly nonlinear finite element problems. To overcome these challenges, the framework integrates the following innovations: 1. Crystal plasticity modeling – A finite-strain, dislocation-density-based model incorporating geometrically necessary dislocations to account for grain size effects. 2. High-performance mesh generation – A robust, fast mesh intersection technique to embed microstructural features into the blade’s finite element mesh. 3. Efficient solver technology – A multipreconditioned domain decomposition solver with a grain-based partitioning strategy for improved computational efficiency and scalability. This framework enables a direct investigation of grain orientation and size effects on plastic strain localization in turbine blades. Furthermore, it facilitates the creation of synthetic databases for statistical analysis and reduced-order modeling. As an illustrative application, we present a graph neural network trained on these high-fidelity simulations, showcasing the potential of this approach for advanced turbine blade design and performance optimization.