The generation of large-scale microstructural models is essential for understanding the process-structure-property relationships of metals. We present a software LEGOMAT, applicable for large-scale real-time integration of microstructural models with millions of grains within an engineering CAD model. We display the use of LEGOMAT software for an additively-manufactured (i.e., laser engineered net shaping) part by embedding microstructures along user-specified laser paths while accounting for hatch widths, layer thicknesses, and scan directions. Here, the surface micrographs are obtained from location-specific EBSD images, which are converted to three-dimensional (3D) representative microstructures using the Markov random field technique. These 3D microstructures are seamlessly merged into an engineering-scale CAD model, using a patch-based optimization process that maps every finite element within the CAD model to a synthetic microstructural domain. Cross-validation of morphological features (e.g., grain size, shape, orientation, two-point correlation), as well as mechanical properties are carried out to validate our reconstruction strategy.