Generative AI with Large Language Models occupies a really useful gap in the market, sitting well above the everyone-welcome conceptual courses but well below a full multi-month specialisation, and it does so with unusual technical honesty. It was built by DeepLearning.AI together with AWS, and the involvement of working engineers shows in how much attention goes to the parts that actually matter in production rather than just the theory. Across roughly three weeks you go through the full lifecycle of a large language model, starting with how the transformer architecture works, moving through prompting and in-context learning, then into the meat of the course which is fine-tuning, parameter-efficient methods, and reinforcement learning from human feedback. What I appreciate most is that it refuses to wave its hands at the hard, unglamorous questions, so you spend real time on compute cost, scaling laws, and the practical trade-offs of getting one of these models to behave and to run affordably, which is exactly the knowledge that separates someone who has read about LLMs from someone who can actually work with them.
The labs are hands-on with real models rather than cartoon examples, and because AWS co-produced it you are working inside their tooling, which is a strength if you live in that ecosystem and a mild bias if you do not. The flip side of all this depth is that the prerequisites are real, and the course says plainly that you need Python and a working knowledge of machine learning, which is true. If you arrive without that footing you will spend the whole time underwater, so this is genuinely not a starting point, it is a second or third step. The density is the other thing to flag, because three weeks is not long for this much substance, and most people will get more from it on a second pass than they do on the first.
You can audit it free and pay the usual Coursera subscription for the certificate and graded work. My verdict is that for a developer or data professional who already has the basics and wants to truly understand how modern language models are trained, tuned, and shipped, this is about the best concentrated course you can take right now.