Scientific modeling concerns a continual search for better models for given data sets. This process can be elegantly captured in a Bayesian inference framework. ForneyLab enables largely automated scientific design loops by deriving fast, analytic algorithms for approximate Bayesian inference.