Carbon nanotubes (CNT) have extraordinary mechanical properties, but to take advantage of these properties in composites, bundles, and ropes requires strong bonding to achieve significant CNT-CNT or CNT-matrix load transfer. This work is a computational study examining strengthening CNT composites and bundles on the quantum, atomistic, and meso- scales. Density functional theory (DFT) and classical molecular dynamics (MD) are used to evaluate methods to improve bonding CNT-matrix crosslinking by the inclusion of dopants, defects, functional groups, and curvature. DFT and MD are also used to quantify the load transfer of CNT-CNT sulfur crosslinks. A coarse-grained (CG) technique for modeling CNTs on the mesoscale is extended to include nonconservative frictional forces which are parameterized to model crosslinking. This extended CG model is then used to predict the mechanical performance of CNT bundles on a much larger scale. In addition to utilizing these traditional computational material science methods, a general approach is developed for applying machine learning (ML) to predict the ground state electron and energy density of an atomistic system.