Traditionally, theory developed for dispersion of asolute in a fluid flow has focused on predicting enhanced diffusion which can be observed, even with smooth fluid flow. A distribution will appear to spread much more rapidly than could be explained as the result of molecular diffusion alone, and decades of theory have been developed to successfully explain this. In our prior work, we developed asymptotic theory which predicts not only spread (variance), but also the sign differences in the distribution's skewness depending on cross section. This theory has had success matching with both experiment and numerical simulation. I will introduce the setting with these past results, then present a computational framework to extend this work past idealized cross sections and explore some questions in shape optimization towards more precise control of the tracer distribution, exploring the ability to design channels to match a given specification.