Metal parts manufactured via the powder bed fusion (PBF) process have drawn tremendous interest in the automotive industry. While numerous studies have shown the unique microstructure of the metal from the PBF process, significant variation of material properties with process parameters has been widely observed, indicating that huge amounts of experiments are required during material characterization. Thus, multiscale material modeling approaches are in great demand so that the properties of the metals via the PBF process can be predicted with confidence, to save costs and time during the design stage. In the present study, a multiscale modeling approach is proposed in which the microscale and mesoscale models are considered in finite element analysis. At the microscale, the model captures the microstructure characteristics within the melt pools to predict the representative properties resulting from epitaxial grain morphology and orientation. The properties are then homogenized and input into a mesoscale model in which the “fish-scale-like” melt pools and boundaries between them are modeled. Stochastic reconstruction of the micro- and mesoscale models are performed based on statistical microstructure information obtained from optical micrographs and scanning electron microscopy (SEM) images. Predicted mechanical properties are compared with experimental data to demonstrate the capability of the approach. The study keeps focus on AlSi10Mg built by selective laser melting (SLM), while universal applicability to other material systems is expected.