I generally recommend that before students take a machine learning course, they should know how to program, and they should have taken somewhat advanced courses in probability and linear algebra. Many professors post their course notes and lecture videos online (as I do), so I would recommend trying to watch those to gain more knowledge.

The best way (in my view) to learn Machine Learning after taking a course is to work on a real problem. There is a lot to learn because real data science is messy: it takes effort to specify the goals correctly (based on what the domain experts say) and to collect and clean the data. It is common to find that the problem has been formulated in a problematic way and needs to be fixed. Important pieces of advice are:

1. Keep in mind that Neural Networks are excellent for computer vision, but they don't tend to give an advantage on many other kinds of data. Thus, it is worthwhile to learn other techniques besides Neural Networks, as they can be difficult to train, and are often black boxes.

2. Interpretable Machine Learning is critical for fairness and trust, but it is a topic that is rarely taught in courses.

3. Ethics are important. Data scientists should evaluate the ethics of the project themselves, not just do what their manager tells them! Ethics is another topic rarely taught in data science courses.