sentdex (YouTube channel)

What the channel covers

sentdex describes itself as Python programming tutorials going further than just the basics: machine learning, finance, data analysis, robotics, web development and game development. The channel has well over a thousand videos, and the companion site pythonprogramming.net hosts text versions of many tutorials. The style is hands-on: Harrison writes code live, explains it as he goes and lets you see the bugs and fixes along the way.

In recent years the channel has leaned towards experiments with robots, local models and home lab hardware, while the older series remain some of the clearest project-based introductions to machine learning in Python.

Playlists worth starting with

• Neural Networks from Scratch in Python: builds a neural network with plain Python and NumPy, layer by layer, so you understand what frameworks do for you.
• Pytorch - Deep learning w/ Python and Deep Learning basics with Python, TensorFlow and Keras: the step up to real frameworks.
• Reinforcement Learning and Reinforcement Learning with Stable Baselines 3: agents that learn by trial and error.
• Data Analysis w/ Python 3 and Pandas: a practical start on data work.
• Learning to program with Python 3 (py 3.7): the beginner series if your Python is shaky.
• Self-driving cars with Carla and Python and Python AI in StarCraft II: longer project series once the basics are solid.

Some older series use library versions that have since changed. Expect to adjust imports and APIs, and treat that as practice.

Who it suits

Students who have finished a first Python course and want to see how real machine learning projects come together, and developers who learn best by building. Complete beginners should start with the Python basics series before the machine learning ones.

How to learn from it without getting stuck in tutorial loops

• Code along once, then rebuild alone. After a series, close the video and rebuild the project from memory. Where you get stuck is what you actually need to study.
• Change one thing. Swap the dataset, the environment or the model, and see what breaks.
• Stop at one series per month. It is tempting to binge new projects; finishing and extending one teaches more.
• Write it up. A short README or blog post explaining your version is good evidence for a portfolio.

Pair it with on Sefism

• Follow the Machine learning roadmap for the order of topics, and keep the machine learning cheat sheet open while you build.
• If your Python is new, start with the Programming foundations roadmap.
• PyTorch: Learn the Basics and the NumPy documentation are the references you will check most.
• Use Google Colab if your laptop has no GPU.

Best for: learners with basic Python who want project-based machine learning and do not mind updating older code.