This echoes my own experience. I made an attempt at andrew Ng's revised machine learning course last year. I could understand well enough at a high level how gradient descent worked but its been 20 years since I did any calculus and the level to which my ability to even do algebra had atrophied greatly. To their credit, they derive the equations for you but I did't see how I was going to be able to really apply any of what I learned when the caclulus "handwaiving" is probably the most crucial step.

I'm a fulltime CTO so finding textbooks that can fill in the gaps and finding endless problem sets to solve was just not going to work. Luckiy, A good friend of mind from hack reactor clued me in to mathacademy. I would argue thats its probably one of the biggest underated resources for getting back in mathematical shape. I've been setting aside an hour a day to just grind through the lessons and problem sets that it throws at me. it uses spaced repetition along with an inital placement test to figure out what you're weak at and just hits you with those problems as you improve.

echoing the sentiment in the article, you'll get better just grinding though different problem sets consistently each day with the occasional metaphorical boss battle. Once you realize that, actually getting better at math is more of a logistical challenge (having to track down skill appropriate problems to cut your teeth on) Mathacademy basically automates that completely for you. I've gone from giving up on ever getting into this machine learnign stuff to looking forward to spending next year taking on deep learning.

PS: not paid by mathacademy.com. just an incredibly pleased custoner

Also PS: didn't realize you worked at math academy. any plans on expanding into physics problems? would LOVE these ideas to delve back into phsyics. (especially circuits.)