NumPy is the worst way of abstracting linear algebra into a programming language. Octave is better in not obfuscating the concepts with OO boilerplate and having actually useful printed matrix output instead of ugly list representations. Even Fortran is better.
In practice, you'll often encounter the code before the underlying math (especially in Machine Learning), so developing intuition from both directions is always helpful.
Btw, if you enjoy pure mathematics, starting from Chapter 4 in the book, I begin defining vectors and the axioms of vector spaces:
https://little-book-of.github.io/linear-algebra/books/en-US/...
numpy is an intuitive way of doing those numerical calculations in a general use programming language. only minus is the missing gong and why commands