About LLMs & RAG
Large language models can write, summarize, and answer questions, but they only know
what was in their training data. Retrieval-augmented generation (RAG) closes that gap by
letting a model look up relevant information before it answers. This course starts with
LLM Fundamentals and Prompt Engineering, then shows
you how to use models from code with Working with LLM APIs.
Next comes the retrieval side, which decides whether your system finds the right
context: Embeddings, Vector Databases,
Document Processing & Chunking, and
Retrieval & Reranking.
With those pieces in place you'll assemble a complete system in
Building a RAG Pipeline, speed up development with
LangChain & LlamaIndex, and go further with
Advanced RAG techniques such as query rewriting, agentic retrieval,
and knowing when fine-tuning is the better choice.
Finally, you'll learn how to measure and protect your system with
Evaluation & Guardrails, and how to run it for real users with
Deployment & Production.
Basic Python is required, and the ideas from Introduction to AI will help. This course
walks through the topic in order, so each module builds directly on the last.
Content
12 modules
01
LLM Fundamentals
How large language models work: tokens, context windows, transformers, and why models sometimes make things up.
Tokens
Transformers
Hallucinations
02
Prompt Engineering
Writing prompts that get reliable results: clear instructions, examples, structured output, and step-by-step reasoning.
Prompting
Few-Shot
Structured Output
03
Working with LLM APIs
Calling hosted models from code: chat requests, system prompts, streaming, function calling, and handling errors and costs.
APIs
Streaming
Function Calling
04
Embeddings
Turning text into vectors that capture meaning, and measuring similarity so a system can find related content.
Embeddings
Similarity
Semantic Search
05
Vector Databases
Storing and searching embeddings at scale with indexes, metadata filters, and tools like Chroma, pgvector, and Pinecone.
Vector Search
Indexing
Metadata Filters
06
Document Processing & Chunking
Loading PDFs, web pages, and other sources, cleaning them, and splitting them into chunks that retrieve well.
Loaders
Chunking
Parsing
07
Retrieval & Reranking
Finding the right context: semantic and keyword search, hybrid retrieval, and rerankers that sort results by relevance.
Hybrid Search
Reranking
Top-k
08
Building a RAG Pipeline
Putting the pieces together: retrieve relevant chunks, build the prompt, generate an answer, and cite the sources.
RAG
Context Building
Citations
09
LangChain & LlamaIndex
Using popular frameworks to build RAG applications faster, and knowing when a framework helps and when it gets in the way.
LangChain
LlamaIndex
Orchestration
10
Advanced RAG
Improving results with query rewriting, multi-step retrieval, and agentic RAG, and deciding when RAG or fine-tuning is the better fit.
Query Rewriting
Agentic RAG
Fine-Tuning
11
Evaluation & Guardrails
Measuring answer quality and retrieval accuracy, reducing hallucinations, and defending against prompt injection.
Evaluation
Faithfulness
Guardrails
12
Deployment & Production
Taking a RAG system live: caching, cost and latency, monitoring, access control, and keeping data private.
Deployment
Monitoring
Security
What You Can Do After This Course
By the end of the curriculum you'll have shipped four real projects and covered the
full path from raw documents to a production RAG system. Concretely, you'll be able to:
Explain how LLMs work: tokens, context windows, and transformers, and why models hallucinate.
Write effective prompts: clear instructions, examples, and structured output you can rely on.
Use LLM APIs in code: chat requests, streaming, and function calling in real applications.
Turn documents into searchable data: embeddings, chunking, and vector databases.
Build a RAG pipeline end to end: retrieve context, generate grounded answers, and cite sources.
Improve retrieval quality: hybrid search, reranking, and query rewriting.
Evaluate and protect your system: measure answer quality, reduce hallucinations, and defend against prompt injection.
Run RAG in production: manage cost, latency, monitoring, and data privacy.
Show proof of skill: a chat-with-your-documents app, a semantic search engine, a RAG chatbot with citations, and a RAG evaluation dashboard.