Welding distortion is usually controlled by clamping techniques that can be tack welds, pre-bending, and tension loading. Side heating or fast cooling can also mitigate the distortion in some applications. In addition to the clamping techniques, process parameters affect the distortion so that if one can control the welding process parameters, an optimized profile of such parameters could alleviate the distortion. It is shown in this paper that the distortion can be mitigated by using an optimized profile of welding current and travelling speed. These profiles keep the power per unit length of welding constant. It is shown that an increasing welding current at the beginning and the end of the welding path on an edge welded bar of Aluminum could result in a bar that is closer to flat compared to the constant welding current. Developing an optimized weld process parameter profile requires a trustable computational model to implement a control problem using a predictive model for distortion in front of the weld pool in order to adjust the welding current and speed. Unlike using a constant welding current for the full path of weld, the path length is divided into several sub-paths. For each of weld sub-path the control problem learns from the previous sub-path and tries to find the new value for the welding current and speed that minimize the distortion using predictive Computational Weld Mechanics (CWM). Final deflections of the bar are also compared between a constant welding current and optimized profile of welding current.