Aerodynamic performances may be optimised by the appropriate tuning of Active Flow Control (AFC) parameters. For the first time, we couple Genetic Algorithms (GA) with an unsteady Reynolds-Averaged Navier-Stokes (RANS) solver using the Spalart-Allmaras (SA) turbulence model to maximise lift and aerodynamic efficiency of an airfoil in stall conditions [1], and then validate the resulting set of optimal Synthetic Jet Actuator (SJA) parameters against well-resolved three-dimensional Large Eddy Simulation (LES). The airfoil considered is the SD7003, at the Reynolds number Re=6×104 and the post-stall angle of attack α=14∘. We find that, although SA-RANS is not quite as accurate as LES, it can still predict macroscopic aggregates such as lift and drag coefficients, provided the free-stream turbulence is prescribed to reasonable values. The sensitivity to free-stream turbulence is found to be particularly critical for SJA cases. Baseline LES simulation agrees well with literature results, while RANS-SA would seem to remain a valid model to a certain degree. For optimally actuated cases, our LES simulation predicts far better performances than obtained by suboptimal SJA LES computations as reported by other authors [2] for the same airfoil, Re and α, which illustrates the applicability and effectiveness of the SJA optimisation technique applied, despite using the less accurate yet computationally faster SA-RANS. The flow topology and wake dynamics of baseline and SJA cases are thoroughly compared to elucidate the mechanism whereby aerodynamic performances are enhanced.