[hackermailman]

You should probably look up the undergrad math courses of the grad school you want to go to, they will have lectures notes and textbook/chapter reading recommendations so structured and you aren't going through enormous reference books. Search the textbook name on youtube, often there will be some lectures for it sometimes even by the author.

There's an ML math prep book https://mml-book.com/ which is basically a crash course, and a series of lectures here for a background in math for an intro machine learning course at CMU https://www.youtube.com/user/professorgeoff though note they aren't as long as they seem, as these were live lectures so they start late, have breaks between exercises, etc. If you've read the Elements of Statistical Learning 2e you likely know much of this already.

Personally my recommendation is go through a Wilberger course, this set of undergrad lectures is for the Stillwell book 'Mathematics and it's History' https://www.youtube.com/playlist?list=PL55C7C83781CF4316 it will intuitively cover differential geometry, topology, group theory, polynomials etc, to the depth of Stillwell's book and if you see something that interests you or that you forgot, then you can pursue it taking formal courses. I'd recommend his Linear Algebra course too on the same youtube channel he uses clear definitions for everything so when you get to abstract 3D vector spaces it makes sense.

Anybody with a complete shit background in math like I used to have try the Wildberger foundations playlists on the same channel watching how he writes proofs, then pick up some large book written by Knuth and start attempting the exercises as a weekly hobby, which will now be possible to do. This is also how you retain these skills by using them on a regular basis, at least for me anyway.