Serratiopeptidase (SRP), a 50 kDa metalloprotease produced from Serratia marcescens species, is a drug with potent anti-inflammatory property. In this study, a powerful statistical design, evolutionary operation (EVOP), was applied to optimize the media composition for SRP production in shake-flask culture of Serratia marcescens NRRL B-23112. Initially, factors such as inoculum size, initial pH, carbon source, and organic nitrogen source were optimized using one factor at a time. The most significant medium components affecting the production of SRP were identified as maltose, soybean meal, and $K_2HPO_4$. The SRP so produced was not found to be dependent on whey protein, but rather was notably induced by most of the organic nitrogen sources used in the study and free from other concomitant protease contaminant, as revealed by protease inhibition study. In addition, experiments were performed using different sets of EVOP design with each factor varied at three levels. The experimental data were analyzed with a standard set of statistical formula. The EVOP-optimized medium, with maltose 4.5%, soybean meal 6.5%, $K_2HPO_4$ 0.8%, and NaCl 0.5% (w/v), gave a SRP production of 7,333 EU/ml, which was 17-fold higher than the unoptimized media. The application of EVOP resulted in significant enhancement of SRP production.