I found code LLMs to be very useful for rewriting nested for-loops in Python as nice vectorized operations in NumPy. One has to be careful about unexpected array materialization, but generally it works really well and saves a lot of time.