A multi-level parameterization approach for shape optimization problems is developed in the framework of Particle Swarm Optimization (PSO) algorithms. The Free-Form Deformation (FFD) technique based on a third order Bézier tensor product is used to construct a set of nested parameterizations. Then, optimization is carried out in the corresponding search spaces of variable dimension using a PSO algorithm. A method is developed to transfer the information collected from the optimization using a given parameterization, to the optimization procedure using finer level. The proposed method is applied to shape optimization problems in electromagnetics and aerodynamics. Improvements of the convergence rate and the fitness obtained are reported.