This has to be one of the easiest open source projects you can contribute to if you are looking to do so.
Each contribution is small and modular, so you don't have to learn several thousand lines of code.
The "real world" knowledge you have to learn is pretty specific and easy to convert into code (compared to, say, legal opinions or object recognition).
But that is for good reason -- it seems to exist mainly for pedagogical reasons. It is more like contributing to a textbook. Contributing to the "real world" software (NumPY, OpenBLAS, etc) which is optimized and battle-hardened is tougher exactly for that reason.
Each contribution is small and modular, so you don't have to learn several thousand lines of code.
The "real world" knowledge you have to learn is pretty specific and easy to convert into code (compared to, say, legal opinions or object recognition).