A computational design optimization environment is developed, handling, for the first time, streamline dies used for profiles in unplasticized polyvinyl chloride (uPVC) having multiple complex features, as well as simpler designs. Die cavity cross sections are described by planar contours, such as the cutting paths for wire electrical discharge machining of the plates from which streamline profile dies are constructed. Contours are parameterized using key points, and by joining the contours with ruled surfaces, the three-dimensional geometry can be reconstructed. For the optimization a developed flow analysis on each die cross section is used with the avoid-cross-flow strategy. Cross sections are partitioned and the die is balanced to obtain the required flow rate through each. A robust and efficient parallel decoupled optimization strategy is developed. In application to a uPVC window profile, five cross sections were optimized. The number of design variables on each ranged from 2 to 46, and the cross section optimizations converged within one to seven cycles. Compared with the work of an experienced designer making manual changes to the computer-aided design model, guided by computational fluid dynamics analyses, the design quality was comparable or better and computational demands similar; however, the time required from the designer was reduced seven times. POLYM. ENG. SCI., 2013. (c) 2012 Society of Plastics Engineers