I've heard only the best things about Andrew Ng's machine learning course but never came around to do it. It is pretty old by now and with the dramatic development in recent times I wonder how relevant it still is?

Apart from the programming exercises being in Matlab (which nobody uses for ML nowadays), the course is still solid. The theoretical concepts and math that the course covers are still relevant and provide a good foundation for someone starting out.

Now it uses Python: https://www.coursera.org/specializations/machine-learning-in...

The good news is that someone converted the matlab/octave exercise templates into numpy, and published them as jupyter notebooks that can work with the course's auto-grader: https://github.com/dibgerge/ml-coursera-python-assignments