Courses

Canonical intro to ML — supervised learning, unsupervised learning, and best practices. The best starting point.

5-course series: neural nets, CNNs, sequence models, hyperparameter tuning, and structuring ML projects.

Rigorous theoretical grounding in RL — MDPs, TD learning, Q-learning, function approximation, and policy gradients.

Stanford's graduate ML course with rigorous mathematical treatment of every major algorithm. Free on YouTube.

The gold standard CV course. CNNs, detection, segmentation, and generative models — all from first principles.

Word vectors, RNNs, attention, Transformers, and LLMs — one of the most thorough NLP courses available.

Graduate-level deep RL: policy gradients, model-based RL, offline RL, and real-world applications.

Builds backprop, MLP, and a GPT from scratch in pure Python. The clearest explanation of how LLMs work.

The definitive linear algebra course. Essential math for understanding ML, PCA, SVD, and neural networks.

MIT's intensive deep learning boot camp. Dense, fast-paced, covers all major architectures in a few lectures.

Top-down, code-first approach. Gets you fine-tuning state-of-the-art models before teaching the math behind them.

Hands-on with the Transformers library — tokenization, fine-tuning BERT/GPT, and deploying NLP pipelines.