Gear researchers and engineers have been seeking systematic approaches to optimize the microgeometry of gear teeth, especially of hypoid gears. Because their tooth surface deviations from their conjugate counterparts (ease-off ) are small (tens to few hundreds of micrometers), their contact properties are quite sensitive to micro-geometry, which has a significant impact on noise, contact pressure distribution, sensitivity to misalignments, and can also affect mechanical efficiency. In gear design, typical objectives are often conflicting in nature. Therefore, ease-off optimization should be formulated and solved as a proper multi-objective optimization problem. Micro-geometry optimization has long been a time-consuming trial-and-error procedure, mostly based on the individual experience of skilled gear practitioners. One of the earliest published contributions to the solution of this problem was given by Litvin [1] with his local synthesis method. Subsequently, methods aimed at improving the contact pattern and the transmission error function through micro-geometry corrections were proposed, e.g. [2, 5, 6]. Nowadays, simulation models for loaded tooth contact analysis (LTCA) enable to accurately predict contact stresses and deformations. In this study, the Hypoid Analysis Program (HAP) developed by Kolivand and Kahraman [9] was used. This work proposes a method for simulation-based multi-objective ease-off optimization. The proposed method was tested on a real face-hobbed hypoid gear set. Three simultaneous objectives were defined: maximization of mechanical efficiency, minimization of loaded transmission errors, and minimization of the maximum contact pressure. Bound constraints on the design variables were imposed as well as a nonlinear constraint aimed at keeping the loaded contact pattern off the tooth edges, inside a predefined allowable contact region. The results show that the proposed method can obtain optimal ease-off topographies that significantly improve the basic design performances. Also, the method is general enough to be employed for geometry optimization of any gear type.