Statistically based experimental designs were applied to optimize the fermentation medium and cultural conditions for the maximization of neutral protease using three agroindustrial residues (cassava pulp, soybean meal, and wheat bran) and Bacillus subtilis DES-59. The Plackett-Burman design was used to evaluate the effects of variables such as the concentration of substrates, initial pH, shaker's rotating speed, temperature, inoculum size, and incubation time. Among the eight parameters, three significant variables (cassava pulp, soybean meal, and inoculum size) were selected for the optimization study, in which a central composite design was used to optimize the concentrations of cassava pulp and soybean meal and inoculum size and investigate the interactive effects of the three variables. The optimal parameters obtained from response surface methodology are 37.78 g/L of cassava pulp, 15 g/L of soybean meal, and 6.5% (v/v) of inoculum size, respectively, resulting in a maximum neutral protease activity of 4107 ± 122 U/mL.