Learn RAG
Scrimba's Pro course on Retrieval-Augmented Generation, taught by Guil Hernandez across about 1.6 hours. It is the core technique for making LLM apps answer from your data instead of making things up.
Quick answer
Learn RAG is Scrimba's intermediate, Pro-tier course on Retrieval-Augmented Generation: roughly 1.6 hours across 22 lessons, taught by Guil Hernandez. You use embeddings and a vector store to ground an LLM's answers in your own documents, which is the standard fix for hallucination and stale knowledge. It is one of the most directly useful techniques on the AI track.
Learn RAG
ProTaught by Guil Hernandez (opens in a new tab)
Use embeddings and a vector store to make LLM apps accurate and grounded in your own data.
View on Scrimba (opens in a new tab)Is it worth your time?
RAG is one of the highest-leverage techniques in applied AI, and this course teaches it at the right depth: enough to understand embeddings and vector stores and to wire a working pipeline, without drowning you in theory. If you build anything where the model must answer from specific, current information, this pays for itself quickly.
The honest caveat is that it assumes the fundamentals. You should already be comfortable calling an LLM and writing JavaScript before you start here, or the embeddings and retrieval steps will feel like magic. It is also focused on RAG specifically, so it is a technique course, not a broad survey.
What you'll learn
The course builds the RAG pipeline end to end. You learn how embeddings turn text into vectors, how a vector store lets you retrieve the most relevant chunks, and how to feed those into the model so its answers stay accurate and grounded. The interactive format means you assemble the pipeline yourself, which is what makes a slightly abstract idea concrete.
Who it's for, and who should skip it
It fits developers building LLM apps that need to answer from a known body of information: documentation assistants, support bots, internal search. If accuracy and grounding matter to your app, this is essential.
Skip it, for now, if you have not done the AI fundamentals; start with Intro to AI Engineering. For a free first look at RAG, Intro to Mistral AI touches on it before you commit to this.
Prerequisites
JavaScript and a working understanding of calling an LLM. The AI Engineering fundamentals are the ideal lead-in. No prior knowledge of embeddings or vector databases is assumed.
Where it fits
This is a core technique course on the AI Engineer Path, best taken after the fundamentals. It underpins many real AI apps and pairs naturally with the agent courses, since grounded retrieval makes agents far more reliable.
Free or Pro
This is a Pro course requiring a Scrimba subscription. Pro also covers the full AI Engineer Path, the challenges, the Discord, and certificates. See current plans for pricing in your region. For a free preview of RAG concepts, Intro to Mistral AI introduces them.
Strengths and limits
What it does well: it teaches a genuinely high-value technique at a practical depth, with a clear instructor and a build-along format.
Where it is limited: it assumes the AI fundamentals and JavaScript, and it is focused on RAG specifically rather than offering a broad survey.
Related courses and comparisons
- Intro to AI Engineering, the fundamentals to take first
- Intro to Mistral AI, a free preview of RAG
- Learn AI Agents, where grounded retrieval makes agents reliable
- Learn Context Engineering, a related approach to keeping apps accurate
No. It is a Scrimba Pro course requiring a subscription. For a free preview of RAG concepts, Intro to Mistral AI introduces them.
Retrieval-Augmented Generation: using embeddings and a vector store to fetch relevant information and feed it to an LLM so its answers stay accurate and grounded.
Yes. The course assumes you can already call an LLM and write JavaScript. Start with Intro to AI Engineering if those are new.
Guil Hernandez, who also teaches Learn to Code with AI and Intro to Dall-E and GPT Vision on Scrimba.
Whenever an app must answer from specific or current information, such as documentation assistants, support bots, or internal search over your own data.