Spend enough time looking at Python programs and packages for machine learning, and you'll notice that the "JIT decorator" pattern is pretty popular. For example, this JAX snippet: import jax.numpy as jnp import jax @jax.jit def add(a, b): return jnp.add(a, b) # Use "add" as a regular Python function ... = add(...) Or the Triton language for writing GPU kernels directly in Python: import triton im