The performance of most manufacturing processes depends on numerous parameters and their interactions. Most of the time, selection of an appropriate set of machining parameters is done based on experience, trial and error, or both. Electro-discharge machining (EDM) process is complicated and random in nature. The large number of parameters and the inherent complexity of removal mechanisms taking place in EDM make it even more difficult to select machining conditions for optimal performance. The objective of this study is to provide information on the relationships between the key input variables and resultant surface roughness and to develop a response model for surface roughness optimization utilizing factorial designs, direction of steepest descent method and response surface methodology (RSM). Experiments were setup and executed to understand the individual and combined impact of factors that included the following input variables: Gap voltage, depth of penetration, electrode type, average current and pulse duration. Six iterations were executed in the direction of steepest descent for minimization of surface roughness. A higher order model is developed that can relate process inputs to response. The results obtained may be used to recommend process setting to improve process robustness and to obtain desired surface roughness.