Monte Carlo radiation transport enables engineers to accurately predict the behavior of nuclear systems. Over the years, this simulation method has been applied to increasingly complex devices worldwide. Historically the modeling of such systems for radiation transport often requires approximations to geometry, limiting the fidelity of results like particle flux, heat generation, activation, etc. which inform design parameters in other engineering domains. Research at UW - Madison has provided the capability for robust Monte Carlo analysis on CAD geom etry, resulting in higher fidelity geometry representations of devices like advanced test reactors, fusion reactor designs, and high energy particle beams. This capability comes at great computational cost, however, resulting in simulation times much slower those using native geometry representations. This talk will focus on increasing the efficiency of CAD-based Monte Carlo simulations on tessellated models, leveraging alternative data structures and recent advancements in CPU architectures to achieve this end.