Kaggle Learn is one of those resources I am almost surprised is free, given how practical and well-made it is. Kaggle is the platform most data scientists know for its competitions and its enormous library of public datasets, and Learn is its teaching arm, a collection of short courses on things like Python, intro to machine learning, intermediate machine learning, pandas, data visualisation, feature engineering, deep learning and even more current topics. The whole thing runs in interactive notebooks on Kaggle itself, so you are writing and running real code from the first few minutes rather than watching someone else do it. The format is the main attraction.
Each course is deliberately small, the kind of thing you can finish in an afternoon, and it focuses on getting you doing useful work quickly. If you have a specific gap, say you have never really learned pandas properly or you want a fast, concrete introduction to feature engineering, these courses are a brilliant way to plug it without committing to a forty hour program. And because it all lives on Kaggle, the moment you finish you are one click away from real datasets and live competitions where you can actually apply what you just learned, which is a much more natural next step than most course platforms offer. The trade-off is the flip side of the same coin.
These courses are short because they are shallow on purpose. They show you how to do things and they are light on why those things work, so you will not come away with a deep grasp of the underlying statistics or the maths, and the deep learning material in particular is more of a taster than a foundation. The completion certificates are not something I would put any weight on at all. None of that is a criticism so much as a description of what this is, a set of fast, practical primers.
My honest recommendation is to use Kaggle Learn exactly as intended, as a launchpad. Pick the specific courses that match a gap you have, blast through them, then immediately go and apply the skill on a real Kaggle dataset or competition, and pair the whole thing with something more theoretical when you want depth. Used that way, it is one of the best free practical resources in data science.