Numpy and pandas have been the backbone of scientific computing and have been ubiquitous in data science/ AIML problems. They helps us to have an analytical and statistical overview of the data involved in the problem. Numpy and panda both helps us in dealing with large dataset, operating on them, having insights on its nature and hidden patterns within it. They can be seen in action from operating on low level tensors in theano to high level neural network frameworks such as keras, tensorflow etc.
Numpy can accomodate data with similar type. It cannot have custom indexes. Pandas can accomodate series with different data types. It can have custom indexes. Therefore, out of the box import of csv, xlsx, binary files are supported in pandas.
I have aggregated few useful/mostly used features of both to have a feel of hands on with them. Please feel free to extend them. I hope it will be useful in your learning too.