I got an BSc, MSc and PhD myself so I usually don't watch the stuff from beginning to end as I've been exposed to much of it... but let's see.
The Princeton Algorithms courses (Coursera) seem good and is quite popular but honestly as long as I don't interview I rarely ever need those.
From HSE (Coursera) I watched parts of Stochastic processes and bayesian methods. I found those interesting because the approach is somehow different and what is presented is rather rare.
I did most of the AI course on Udacity with Norvig and Thrun. I really liked that one. At university we also followed the AIAMA textbook so also much repetition but I found the presentstion is good.
Probabilistic graphical models (Daphne Koller, Stanford, Coursera) : a classic but honestly I never brought up enough motivation to keep at it for longer.
The deeplearning.ai courses: Watched all the videos from all parts. Think they were OK. Did exercises in the beginning but didn't feel to learn much from them. Too much just transcribing equations and as I work in ML anyway I rather just do my work.
fast.ai: many love it, I never liked them. Too much time wasted with showing how to set up stuff, work with jupyter Notebooks etc. Some hidden gems but have to endure lots of noise to find them.
Probability - The Science of Uncertainty and Data (MIT, edx) seemed great but too much repetition to really do it all.
Udacity Operating Systems/Advanced Operating Systems. Liked them, as I started teaching an OS course at a local institution I used them as refresher.
Udacity intro to parallel computing was great to get started with CUDA.
Udacity. Interactive 3D graphics was also well made but I think it's over 5 years now that I did it, might not be well maintained.
Non CS I like:
https://www.edx.org/course/introduction-to-biology-the-secret-of-life-3
https://www.coursera.org/specializations/immunology
https://www.coursera.org/learn/medical-neuroscience
https://www.edx.org/course/science-cooking-from-haute-cuisine-to-soft-matter