Thanks for this article.

From my point of view and personal experience you are missing one very important point ... theoretical courses at university will not help you too much for practical usecases as a Machine Learning Engineer in most work scenarios.

The kind of data and problems you are presented to as a student do not cover how things will become in real life. Therefore the so called side projects are the most important ones. And again, you should not take a given dataset all the time but maybe create your own one. Draw the bounding boxes yourself, set up the environment on your own without copying a kaggle notebook or google colab. This will you show you exactly where you have to improve and what you need to learn. you can than simply follow some medium tutorials or coursera courses to refresh / improve your personal knowledge for this fields.

Understanding ml models mathematically is only required to a certain amount as a practical machine learning engneer working in cooperations. The only time I see that being a thing is when applying for research jobs in ml, like at the google research lab, but that is whole other story I guess ^^

Cooprations may require a master's degree in computer sicence or similar for certain ml jobs, but in the end they may end up more interested in you if you can present them a complete github repo where you showed that you can build a ml solution of your interest from start to finish, because that is showing them that you will add value to their team imediately.