The most respected free AI courses in the world — from Stanford and MIT to Fast.ai and Karpathy. Use these if you want the math, the intuition, and the long-term foundation.
If you only do four of these, do these four.
Andrew Ng · DeepLearning.AI & Stanford · Coursera
The 2022 remake of Ng's legendary course. Three sub-courses cover regression, classification, neural nets, recommender systems, and RL.
Andrew Ng · DeepLearning.AI · Coursera
Five courses on neural nets, hyperparameter tuning, CNNs, sequence models. Still the most-recommended deep learning curriculum.
Fast.ai · Jeremy Howard
Top-down course that has you training real models in lesson 1. Famously effective for engineers who want to ship, not derive.
Andrej Karpathy · YouTube
Build neural nets from scratch in Python, then build a GPT in pure code. The clearest explanation of transformers anywhere.
Lectures, problem sets and slides published free by the universities themselves.
Andrew Ng et al · Stanford
The full graduate-level ML course — lectures on YouTube, notes and problem sets on the Stanford site.
MIT
MIT's flagship intro to deep learning. Current materials cover generative AI, diffusion, and LLMs.
Chris Manning · Stanford
The canonical NLP/Transformers course — embeddings, attention, transformers, large language models.
Stanford
Guest lectures from leading researchers on transformer architectures, scaling, multi-modality, and reasoning models.
Carnegie Mellon
Louis-Philippe Morency's well-regarded course on multimodal AI — vision-language, audio-language, embodied agents.
FSDL · Sergey Karayev, Josh Tobin
The course that teaches the parts most academic ML courses skip — labeling, deployment, monitoring, ML systems design.
For the working engineer who specifically wants to go deep on language models.
Hugging Face
Free, hands-on course on transformers, fine-tuning, RLHF, and using the Hugging Face ecosystem. Notebooks throughout.
Hugging Face
Modern follow-on covering large language models, RAG, function calling, and building agents with the Hugging Face stack.