Sophomore computer science student, looking at ISLP vs ESL vs mlcourse.ai

For background, I am currently a computer science sophomore, with intermediate skills in Python and C++. I have taken university courses on data structure and algorithms, calc 1-3, linear algebra, and an introductory stat course (which covered confidence interval, Z and T sample test, and hypothesis testing). I also have read up to Chapter 5 of the MML book and am currently self-studying probability theory (through STAT 110 video and textbook by Joe Blitzstein).
I have done a few beginner ML projects with Tensorflow and scikit-learn, but most of the work is in EDA and feature engineering, while the ML model is just a black box that I plug and chug. So now, I want to learn how to implement ML models from scratch. I've been skimming over ISLP, which many people online recommended, but it seems that while it talks about mathematical equations used, I don't really get to implement it; as the labs are a lot of importing an already implemented model then plug and chug.
So now, I am looking at ESL, which I believe is the more detailed and mathematically rigorous version of ISL. However, there aren't any labs or code along to ease beginners in (which I somewhat understand given the intended audience of the book).
Another option I am looking at is mlcourse.ai, which seems to cover mathematics and has some lab/code along for it. But it doesn't seem to span as many subjects as ESL does.
Given these options, I am unsure of which one to pick, should I first finish my self-study on probability theory and then Chapters 6-8 of MML? Then should I do ISLP first or just get into ESL? Or maybe I should do mlcourse.ai first then into ESL? Or should I just do the ML course/book along with the maths? In addition, there is also the data science + feature engineering stuff which I wonder if I should study more about.
Sorry if this seems like a mess, there are just so many things to ML that I am kinda overwhelmed.