Python is an extremely expressive language, that lets you prototype and experiment very rapidly. In many cases, with some cleanups, refactorings, and calls to C modules like numpy, you can convert those prototypes into sufficiently performant real implementations.

Here's a recent write-up of mine where almost all of the time is spent in C extensions but where it was much faster to develop and code in python: https://www.jefftk.com/p/sharding-the-brigade