Statically typed functional programming helps by allowing more properties to be expressed and checked by the program automatically. In previous work, we created the Hasktorch library to research and extend the ability of typed functional programming to implement machine learning systems. Recent work on Hasktorch has focused on a deep overhaul of metaprogramming facilities to produce foreign function bindings of PyTorch’s recently-released, consolidated C++ API known as libtorch. We improved the implementation of shared Haskell tooling for C++ bindings and generated the bindings automatically support thousands of PyTorch C++ functions.