A simulation-based numerical optimization method for double-clad fibres (DCFs) is presented. The technique seeks to determine the cross sectional design that offers the strongest absorption. An encoding key that transforms the twodimensional cross section into a single point in a multi-dimensional space has been developed. The coordinates of the referred point are treated as the variables of an equation-free objective function and the Nelder-Mead algorithm is deployed to search for the function minimizer which is not always found but the function is always minimized substantially. The coordinates of the minimizer or the point delivering the best performance amongst the simplex vertices is decoded to unfold the optimized cross section. During the optimization process, the evaluation of the objective function is treated as a moving boundaries problem and is addressed via an efficient numerical simulation technique operating in the geometrical optics domain. Various optimized cross section designs are generated and compared. The optimum designs are those of a spiral inner cladding with an offset core and a circular one with four embedded birefringence rods with optimized offsets. The concept of improving the multidimensional landscape instead of improving the direct search method itself increases the agility of the polytope and generates more imaginative cross sections with improved absorption characteristics.