Selective Laser Melting (SLM) enables rapid manufacturing of highly customized and complex geometries but is currently limited by difficulties in predicting the final material properties. SLM is characterized by rapid phase changes, large thermal gradients, discontinuities as well as varying length and time scales. These challenges can only be addressed computationally by utilizing a bottom-up multiscale framework consisting of different numerical methods, each operating on their respective scales. Our aim is to provide a consistent connection between micro-scale simulations with Molecular Dynamics (MD) and macro-scale methods such as the Finite Element Method (FEM). Considering length scales from 1 μm to 100 μm, the most prominent structural feature is the single crystal grain. SLM process parameters such as e.g. heat source energy density, scan speed and cooling rates do indeed influence crystal grain shape and lattice orientation, which in turn affect the bulk material parameters such as hardness and ductility at the meso scale. For computational modelling, we rely on Peridynamics (PD) due to its close relation to MD (both are non-local methods). Upscaling MD to PD was already considered in [3] where both models yield corresponding dispersion relations. PD provides the possibility of significant computational speed-up compared to MD. Expanding upon [1], where grains naturally arose due to intentional material instabilities, we present a PD material law where single crystal grains grow from a super-cooled liquid phase based on spontaneous nucleation. Post- solidification processes such as solid-state transformation which is necessary for modelling heat treatment (e.g. annealing) are considered. We show that the grain structure arising from our method matches real world observations and MD predictions and present a coarse scaling methodology build upon the Crystal Plasticity Peridynamics (CPPD) models considered in [2] and [4] that rigorously leads to observed macroscopic bulk material response.