Sentdex, the channel of Harrison Kinsley, is one of those names that anyone who taught themselves Python or machine learning in the last decade will recognise instantly, because for a very long time it was one of the few places offering serious, code first tutorials for free. The catalogue is genuinely vast, covering core Python, data analysis with Pandas, machine learning fundamentals, deep learning with TensorFlow and PyTorch, reinforcement learning, computer vision, natural language processing and a long tail of quirky project series like teaching a model to play games. The defining quality throughout is that you are always in the editor, writing and running real code, rather than watching abstract explanations, and that hands on bias is exactly what a lot of learners need. His Neural Networks from Scratch material deserves a special mention, because building the whole thing in raw Python rather than reaching for a library is one of the best ways to actually understand what is going on inside a network, and he teaches it with real patience.
The honest weakness is age. Because so much of the value here was created years ago, a meaningful chunk of the tutorials rely on library versions that have since moved on, and following them line for line can lead to errors that a newer learner will not know how to fix. The channel is also broad rather than deep, hopping between many topics without a single guiding curriculum, so it rewards people who already have a rough sense of what they want to learn and can assemble their own path, while leaving true beginners a little unsure where to start. The style is deliberately relaxed and unpolished, which I personally like because it feels like pair programming with a knowledgeable friend, but it will not suit everyone.
My honest take is that Sentdex remains one of the most valuable free teachers in this field, and the learn by building approach is exactly right, but you should treat it as a rich library to draw from rather than a modern, structured course. Watch with an eye on the upload date, expect to do some translation to current tooling, and pair it with something more current for the fastest moving topics.