GooFit is a massively-parallel function evaluation framework, written using Thrust for CUDA and OpenMP, designed for doing maximum-likelihood fits with a syntax similar to that of RooFit. It is especially well-suited for doing unbinned amplitude analyses of large samples of events. It can also be used to fit histograms.  Python bindings exist for most of GooFit's methods. We would like to understand what features/functionality users want and what sort of documentation/training would be valuable.