In this study, we present an improved genetic algorithm (GA) to design userspecified filters housed in a waveguide of arbitrary cross-sectional dimensions. An edgebased finite element method (FEM) is employed as the forward solver for the problem. Additionally, the structures generated via the GA are easily constructed using standard printed circuit board fabrication techniques. Two of the major improvements to the GA are: (1) a technique for enhancing the GA's ability to avoid local minima and (2) a fine-tuning mechanism which allows the GA to more efficiently seek out a minimum once a low error has been obtained. Using the aforementioned techniques, numerical/experimental results are presented for notch, low pass, high pass, and bandpass filters.