I personally spoke with several programmers who wanted to "study physics". I showed them a copy of COMSOL multiphysics, explained to them that even as trained physicist with years of experience I could probably understand well only 20% of the problems the software can solve, and clarified that the portion of physics you touch on software like that is the tiny tiny part of physics that is used heavily in engineering tasks, maybe 1% of it, the most boring one. All of them without exception backed down.
The second thing you have to understand is that studying physics beyond the college level requires an analytical depth of reasoning that is not usually reached in software engineering and that there are no easy didactical books (in recent years the situation improved a lot though). So you either have to understand difficult stuff or you need a good in person teacher. Not to mention that only the mismatch between different conventions and schools of thought is gonna cost you a lot of your time if you rely on a sparse set of books. It happens a lot that different books use different space/time/unit/normalization/coordinate basis choices and that means you simply cannot copy and paste formulas without understanding their derivation.
That said I would suggest
- going through the exercises of a college level book like Halliday Resnick. Walter Lewin lectures if you can find them are good too.
- watch YouTube videos, some of them are really well made these days
- the theoretical minimum from Leonard susskind is a good set of lectures for advanced theoretical concept in minimal format for people who have some STEM background. A good series of books in the same style is the no-nonsense series from Jakob Schwichtenberg.
- go through the kind of stuff you do in computational physics. Simulations like you find in finite elements solvers are great visually. A lot of other techniques exist for modeling quantum systems, gravitational systems, statistical mechanics... Engineering-level stuff will have good software to back it (Synopsis for optics and semiconductors, Comsol for a holistic approach, Ansys for statics, Cst for antennas, Lumerical for optics,...). Most of those will have open source equivalents of significantly lesser quality. More research level stuff, like from DFT and beyond, will have open source packages that get sparser and sparser until you get to some github repository that a graduate student maintained for their thesis.