When interblade coupling is weak, the dynamic response of a bladed disk is very sensitive to the presence of uncertainties. Excessive response variation can be very harmful. Previous studies have indicated that introducing blade-to-blade difference in nominal design, known as intentional mistuning, could reduce the level of response variation. In this research, an efficient computational framework that yields the optimal design of intentional mistuning is developed to maximize the bladed disk reliability. Both the random uncertainty of blades and the interval uncertainty of disk connections are considered. The Metropolis-Hastings algorithm is applied to find the worst case response under interval uncertainty, and Monte Carlo simulation is employed to account for the random mistuning effect. A gradient-based approach is then established to find the minimum design modification needed to achieve a designated reliability level. Case studies are carried out to illustrate the effectiveness of the proposed method.