Nanotherapy is an emerging cancer treatment that relies on nanoparticles to carry cytotoxic drugs to tumors without systemic exposure to the drugs. The design of the drug-loaded nanoparticles highly dictates their intratumoral biodistribution. However, previous studies have primarily relied on empirical selection of nanoparticle designs, which is constrained by financial and temporal limitations, leading to suboptimal efficacy of the treatment. In this work, we apply numerical optimization to generate optimal nanoparticle designs for cancer nanotherapy. We use two mechanistic models to evaluate treatment therapeutical and toxicological parameters. We then formulate single and bi-objective optimization problems that systematically simulate the mechanistic models at different candidate nanoparticle designs to optimize nanoparticle accumulation at the tumor, percent of dead tumor area, and tumor diameter after treatment, as well as to quantify tradeoffs between the competing objective functions. Since the gradients of the objective functions evaluated using the mechanistic models may be unavailable or cannot be approximated reliably, we employ derivative-free direct search optimization algorithms supported by convergence theory. Results revealed design solutions consisting of nanoparticle size, aspect ratio, surface ligand density, and drug diffusion coefficient. The presented optimization framework suggested new treatment protocols and generated new hypotheses that motivates future experimental investigations. This framework can be easily adapted to other cancer therapy modalities.