Probabilistic programming languages (PPLs) allow one to construct statistical models to describe certain problem domains, and then simulate data or perform inference over them. In many PPLs, models lack reusability as they are forced to be defined for a specific use-case: simulation or inference. In other PPLs, models lack the ability to be combined or composed. We describe a DSL in Haskell for modularly defining probabilistic models which are combinable and composable, and can be reused for both simulation and inference.