Particulate composite materials have a broad range of potential applications in engineering and other disciplines. Accurate modeling of their microstructures and fast generation of the finite element meshes play a vital role in investigating many micromechanical phenomena and improving understanding of the underlying failure mechanisms. Due to the exceedingly intricate multiscale internal structures that they possess, the modeling and meshing of their microstructures still remain difficult in general. In this work, we present a computational framework and methodology for the representation, simulation, and mesh generation of 3D stochastic microstructures of particulate composites. Towards this goal, we propose a multi-level multiscale scheme that allows for capturing the multiscale structures of particulate composite materials at both the coarse and fine scales. A briging scale approach based on heat kernel smoothing is also presented to seamlessly link the coarse and fine scales. In addition to the microstructural modeling of particulate composite materials, we also develop an adaptive curvature-based surface and volume mesh generation algorithm for particulate composite microstructures with complex surface texture. Following the implementation of the morphology and mesh generation algorithm, a series of numerical examples are presented to demonstrate the capability and potential of the proposed method.