Restarted doing coding problems after a break and need some further guidance.

I started doing Leetcode problems again after a year of nothing. I don't know how much of this would help me. I am not a CS major. I am currently working as a ML Consultant in a Fortune 500 company. I

I started doing Leetcode problems again after a year of nothing. I don't know how much of this would help me. I am not a CS major. I am currently working as a ML Consultant in a Fortune 500 company. It pays good but want to switch to Research roles in MAANG. I did a physics major and AI master's degree. I used to do Leetcode a year ago and lost all my confidence because I could not solve them properly. I touched it again today, did proper studying and gained some confidence after solving 10 of Interview-150 problems in 2 hours. This can be the momentum I need to start solving more and getting better. Looking back, I think my preparation last time was unfocused. I relied on too many resources, which made things confusing. Fortunately, the company I joined did not emphasize DSA much and instead focused directly on ML/AI in the interview. I realized that my previous preparation approach was not very effective and I am planning to do it my way instead of following a specific cheatsheet. Now, if I want to move to research engineer roles especially in AI Safety, AI Engineering, LLM evals, how much Leetcode/DSA Knowledge should I prepare? What other things should I keep in mind? Do Cheatsheets and roadmap actually work or did they not work for me specifically?