This study deals with the use of optimization algorithms to determine efficient param- eters of flow control devices. To improve the performance of systems characterized by detached flows and vortex shedding, the use of flow control devices such as oscillatory jets, are intensively studied, using numerical as well as experimental methods. However, the determination of effi- cient control parameters is still a bottleneck for industrial problems. Therefore, we propose to couple a global optimization algorithm with an unsteady flow simulation to derive efficient flow control rules. We consider as testcase the turbulent flow over a backward facing step, including a syn- thetic jet actuator. The aim is to reduce the time-averaged recirculation length behind the step by optimizing the jet blowing/suction amplitude and frequency. The Unsteady Reynolds- Averaged Navier-Stokes (URANS) equations are solved within a Mixed finite-Element/finite- Volume (MEV) framework using the near-wall low-Reynolds number one-equation Spalart- Allmaras turbulence closure. The steady flow simulation without control is first validated by comparison with experimental and numerical data. Then, the optimization method EGO (Effi- cient Global Optimization), based on the construction of a Gaussian surrogate model, is cou- pled with the solver and applied to the unsteady flow with actuation. It is shown that the time- averaged recirculation length can be shortened when suitable control parameters are used.