Setting the extruder operating conditions, or establishing the adequate geometry of the screw, for a given group of specifications, is dealt with as an optimization problem where the solution is searched in a nonconvex space. In practice, this is a conflicting multi-attribute optimization problem, where various optima may coexist. The methodology developed involves the maximization of an objective function, quantifying the adequacy of the extruder-die combination response to particular inputs, whose values are computed using an extrusion modeling package developed for this work. Genetic Algorithms (GAs) are used to generate better sets of inputs. In a second stage, multi-objective optimization through Pareto curves, coupled to GAs, is implemented since, beyond optimization, it also provides a better understanding of the characteristics of the extrusion system under study.