Retrieval Augmented Generation (RAG)
by DeepLearning.AI
A full-length RAG course that builds every stage yourself, then takes it to production.
Overview
Retrieval Augmented Generation (RAG) is DeepLearning.AI's full-length RAG course — roughly 26 hours across five modules, with nine code examples and ten graded assignments — taught by Zain Hasan, a Senior AI/ML Developer Relations Engineer at Together.ai who also lectures at the University of Toronto. It is deliberately not a framework tutorial: you build every stage of the system so the design decisions stay visible. Module one covers the three components (LLM, knowledge base, retriever) and has you ship a working system before the theory arrives. Module two is retrieval proper — keyword search, semantic search, metadata filtering, and combining them into a hybrid pipeline with evaluation attached from the start. Module three scales retrieval to vector databases and covers the parts that actually decide real-world quality: chunking strategy, query parsing and reranking. Module four moves to the generation half: prompting over retrieved context, hallucination detection, agentic retrieval loops, and when finetuning beats retrieval. Module five is production — monitoring, evaluation, latency optimization, security, and the cost-versus-quality tradeoff. The framing throughout is tradeoffs rather than one recommended stack, which is what lets the material survive tool churn. On Coursera it holds a 4.8 rating from 212 reviews across roughly 54,000 enrollments. Intermediate Python is assumed; the first module is free to audit, and full access plus the certificate require a subscription.
At a Glance
- Topic
- RAG
- Level
- Intermediate
- Format
- Course
- Cost
- Freemium
- Duration
- ~26 hours, self-paced across 5 modules (Coursera pacing: 3 weeks at ~10h/week)
- Provider
- DeepLearning.AI
- Hands-on
- Yes — code/exercises
- Certificate
- Available
What You’ll Learn
- ✓Build a complete RAG system end to end, from retriever through to generated answer
- ✓Implement keyword, semantic and hybrid search and measure which wins on your data
- ✓Choose chunking strategies and query parsing that measurably change retrieval quality
- ✓Apply reranking to improve the precision of documents handed to the model
- ✓Detect and mitigate hallucination in answers that are supposedly grounded on retrieved text
- ✓Decide when agentic retrieval or finetuning beats a straightforward RAG pipeline
- ✓Monitor, evaluate and cost a RAG deployment running under production load
Highlights
- •Full-length at roughly 26 hours rather than a one-hour short course, and it reaches production
- •Taught by a developer relations engineer at Together.ai who also lectures at the University of Toronto
- •Frames every stage as a cost, speed and quality tradeoff instead of prescribing one fixed stack
- •Ten graded assignments and nine code examples, so the retrieval work is genuinely hands-on
- •Rated 4.8 from 212 reviews across roughly 54,000 Coursera enrollments
Who It’s For
Best For
- ✓Engineers building document Q&A or internal search over proprietary data
- ✓AI engineers who have wired up a RAG demo and now need it to survive production
- ✓Teams evaluating vector databases and retrieval strategies before committing to one
Prerequisites
- •Intermediate Python — you will be writing retrieval and evaluation code
- •Basic generative AI familiarity: embeddings, prompting, what an LLM call costs
- •High-school level math is sufficient; no linear algebra background needed
FAQ
What is Retrieval Augmented Generation (RAG)?
DeepLearning.AI's full-length RAG course, taught by Zain Hasan of Together.ai across five modules with nine code examples and ten graded assignments. It covers keyword, semantic and hybrid search, chunking, query parsing, reranking, hallucination detection and production monitoring. Afterwards you can design a RAG system for real requirements and defend each retrieval and generation choice on cost, latency and quality.
Is Retrieval Augmented Generation (RAG) free?
Retrieval Augmented Generation (RAG) offers free content, with paid options for certificates or premium features.
What level is Retrieval Augmented Generation (RAG) for?
Retrieval Augmented Generation (RAG) is aimed at a intermediate audience. Recommended background: Intermediate Python — you will be writing retrieval and evaluation code, Basic generative AI familiarity: embeddings, prompting, what an LLM call costs, High-school level math is sufficient; no linear algebra background needed.
How long does Retrieval Augmented Generation (RAG) take?
Expect roughly ~26 hours, self-paced across 5 modules (Coursera pacing: 3 weeks at ~10h/week). Most learners work through it at their own pace.
What will I learn from Retrieval Augmented Generation (RAG)?
You'll learn: Build a complete RAG system end to end, from retriever through to generated answer; Implement keyword, semantic and hybrid search and measure which wins on your data; Choose chunking strategies and query parsing that measurably change retrieval quality; Apply reranking to improve the precision of documents handed to the model; Detect and mitigate hallucination in answers that are supposedly grounded on retrieved text; Decide when agentic retrieval or finetuning beats a straightforward RAG pipeline; Monitor, evaluate and cost a RAG deployment running under production load.
Topics
Sources
This page was written from 2 sources, 1 on domains other than deeplearning.ai.