A design tool incorporating classical sonic boom theory, computational fluid dynamics and a multi-objective genetic algorithm was developed for low-boom supersonic aircraft optimization. Both sonic boom and drag were optimized simultaneously and a Pareto optimal set of designs ranging from minimum boom to minimum drag was obtained for each optimization. Since sonic boom was optimized directly, the method had broader applicability than the traditional inverse method. A high-order three-dimensional panel method was used for sonic boom prediction. The traditional linear source model was fast but did not account for wing-body aerodynamic interaction. Euler solutions were expensive for computing sonic booms because a large number of grid points were needed to accurately predict the pressure signature away from the aircraft. For the Mach number and configurations of interest, the panel code showed good agreement with Euler but at a fraction of the cost. A loudness metric was shown to have advantages over initial overpressure and peak overpressure for measuring shaped sonic booms. However, optimization results obtained using calculated loudness raised concerns about the robustness of the solution to atmospheric disturbance. Peak overpressure minimization also produced reasonable sonic boom signatures and appeared more robust to atmospheric uncertainties, but the resulting loudness was not as good. Better convergence was also observed with peak overpressure. Two supersonic business jets were optimized. One was a conventional configuration; the other was a natural laminar flow configuration. Optimization results obtained using loudness and peak overpressure were compared. A non-axisymmetric fuselage was optimized and found to reduce the inviscid drag by 9 to 30 percent at the same sonic boom loudness.