A computationally efficient procedure, NLPANOPT, is developed for the preliminary design of minimum-weight thin-walled stiffened composite panels loaded in uniaxial compression based on a geometrically nonlinear analysis. An approximate, semi-analytical nonlinear analysis code, NLPAN, which requires buckling eigenfunction information from the buckling analysis code, VIPASA, is linked with the optimization code ADS. A blade-stiffened and T-stiffened panel are designed for specified loads using NLPANOPT for postbuckling response and PASCO for buckling-critical response. Comparisons of panel weight and imperfection sensitivity between the NLPANOPT designs and PASCO designs are presented. In general, the designs obtained with NLPANOPT are lighter and less imperfection sensitive than the designs obtained with PASCO. The nonlinear analysis allows for a more accurate prediction of the true strength of the stiffened structure, by accounting for postbuckling strength and modal interaction. The effect of laminate stacking sequence is also investigated. The current design procedure requires the stacking sequence to be prescribed, proving to be a limitation in the design procedure. / Master of Science