An overview of the compiler and runtime pipeline that Numba uses to take Python source code, turn it into a CUDA kernel, and launch it. It pulls in a variety of Numba internals to illustrate how the different parts of the pipeline work.
See the notebook Life of a Numba Kernel with output