Bacillus subtilis C4 was isolated from waste water of a Thai-silk dyeing factory, and identified as a potent strain for producing silk-degumming protease. To optimize the protease production of this bacterial strain, seven fermentation variables were screened using a Plackett-Burman design, and were then further optimized via response surface methodology based on a central composite design. Three significant variables, i.e., soy flour, skimmed milk, and shaker speed, were selected. The optimal values were 2.0% soy flour, 0.1% skimmed milk, and a shaker speed of 280 rpm. The experimental result (1537 units/ml) in a medium optimized for protease production was in good agreement with the predicted value of a quadratic model (1576 units/ml), thus confirming its validity. In addition, the adequacy of the model was supported by a coefficient of determination (R 2 ) of 0.912. Protease production in the optimized medium (1537 units/ml) increased 2.2-fold over that of the non-optimized medium (729 units/ml) in the shaken flask culture. When the experiment was scaled up in a stirred tank reactor, 1891 units/ml protease activity was achieved at 27 h of cultivation, which was an overall 2.6-fold increase over the basal medium.