I love how it's possible to do anything that comes to mind when someone says "Math" with NumPy library. It contains so many functionality to read, manipulate, calculate, visualize data. It provides a fundamental base, almost a platform to perform everything. One can create a simple logistic regression algorithm from the scratch, or a complex deep neural network with the same tools, train it, optimize it. Nevertheless the tools Data Scientists use are already built on NumPy, for example: Pandas, Sci-kit Learn. Not to mention it's efficient to the extreme. Since the functions exist in NumPy are half written in C and are vectorized implementations, they are tens of times faster than writing for loops in Python. Linear algebra operations are especially critical in this term since implementation of machine learning algorithms, especially neural networks need vectorized, fast implementations.

Numpy is an amazing library, the best thing I like about Numpy is it's performance. Numpy is very very faster as compared to Python lists. They have built in array data structure, which is really easy to work with and faster. In Numpy array the matrix multiplication and vector manipulation is super fast. Overall it is best library for Machine Learning related stuff, research related work

I used this library in an online Python course. We didn't go too deep into NumPy, but we used it to convert images to arrays for computer vision applications. Given that NumPy was designed for scientific computation and deep learning, I'm really impressed at its versatility in other areas such as computer vision.