I think a better system is to regularly study fundamentals and always deep dive in your work. Case in point, I've never had problems passing interviews by FAANG or other hot startups, I always got title bump when switching jobs, and I never spent extra time prepping for interviews. Say you're an ML engineer who builds an image recognition pipeline using Spark and PyTorch. Do not just be content with assembling a number of open-source solutions to make your pipeline work. Instead, study the internals of Spark, understand the math and algorithms behind your image recognition models, read survey papers to understand the landscape of data processing and image recognition or further, machine learning, and implement a few models and try to optimize them. Similarly, if you work on database systems, do not just stop at being familiar with MySQL or Postgres or whatever. Instead, understand how transactions work, what consistency means, how principles of distributed systems play out. Study Jim Gray, Gottfried Vossen, Maurice Herlihy, Leslie Lamport, the database red book and its references... You get the idea.
As for leetcoding, replace it with study of algorithm designs. Study Jon Kleinberg's book or Knuth's writings (no, you don't have to read through his books, but his writings are incredibly insightful even for mortals like us), for instance. Instead of working out hundreds of back-tracing problems, study backtracking's general forms.
People tend to underestimate the effect of regular study for years. You'll find that in a few years your knowledge will converge and you will be able to spend less time to incrementally improve your skills, and you will have so many concepts to connect to greatly benefit your projects.