USAIN BOLTZ is a fast Python library for the uniform random generation of tree-like structures. It allows the user to specify both (1) the data structure they wish to sample, using simple combinators similar to those of context-free grammars, and (2) their memory representation. The underlying algorithms are optimised Boltzmann samplers allowing to get approximatesize uniform random generation in linear time. Experimental results show that USAIN BOLTZ matches the performance of the experimental Arbogen package for OCaml, and out-performs the Boltzmann brain Haskell library, while being easier to integrate into existing scientific tools such as Sagemath.