This work deals with the problem of the least-weight design of a composite stiffened panel subject to constraints of different nature (mechanical, geometrical and manufacturability requirements). To face this problem, a multi-scale two-level (MS2L) design methodology is proposed. This approach aims at optimising simultaneously both geometrical and mechanical parameters for skin and stiffeners at each characteristic scale (mesoscopic and macroscopic ones). In this background, at the first level (macroscopic scale) the goal is to find the optimum value of geometric and mechanical design variables of the panel minimising its mass and meeting the set of imposed constraints. The second-level problem focuses on the laminate mesoscopic scale and aims at finding at least one stacking sequence (for each laminate composing the panel) meeting the geometrical and material parameters provided by the first-level problem. The MS2L optimisation approach is based on the polar formalism to describe the macroscopic behaviour of the composites and on a special genetic algorithm to perform optimisation calculations. The quality of the optimum configurations is investigated, a posteriori, through a refined finite element model of the stiffened panel making use of elements with different kinematics and accuracy in the framework of the Carrera's Unified Formulation (CUF).