Learn Retrieval Augmented Generation

Learn Retrieval Augmented Generation is an interactive Python course on Boot.dev, taught by Isaac Flath, previously head of data science consulting at Centriam, and announced in Boot.dev's November 2025 newsletter (published 30 October 2025). It runs to 75 lessons and about 40 hours across 12 chapters, and instead of assembling a framework it has you implement each search technique yourself. The chapters progress in order: Preprocessing (normalizing and cleaning a raw corpus), TF-IDF (inverted indexes and weighting schemes), Keyword Search (BM25 refinements), Semantic Search (embeddings, similarity metrics and vector databases), Chunking (context-preserving segments), Hybrid Search (blending lexical and semantic scores), LLMs (query expansion), Reranking (re-scoring retrieved candidates), Evaluation (precision, recall and relevance), Augmented Generation (feeding retrieved context to an LLM), Agentic (autonomous agents) and Multimodal (extending RAG to images and other modalities). The running project is Webflyx, a movie search system built as command-line Python scripts over a local movie dataset, set up with uv and argparse; LLM calls go through OpenRouter using the OpenAI SDK with free models. The course is freemium: a free account opens some chapters, a membership unlocks the rest, and a certificate of completion is offered. On Boot.dev it is rated 4.7 out of 5 from 145 reviews, with 3,879 students enrolled. An independent 2026 roundup reports reviewers calling it one of the best fundamental RAG courses, but notes that it is Python-only, that some lessons are computationally heavy, and that building primitives from scratch is slower if you only need a deployable pipeline.

Topic

RAG

Level

Intermediate

Format

Course

Cost

Freemium

Duration

~40 hours, self-paced (75 lessons across 12 chapters)

Provider

Boot.dev (Isaac Flath)

Hands-on

Yes — code/exercises

Certificate

Available

• ✓Normalize and clean a raw text corpus so it is ready for indexing
• ✓Build an inverted index and a TF-IDF weighting scheme for keyword search from scratch
• ✓Tune keyword retrieval with BM25 refinements to improve the quality of ranked results
• ✓Apply vector embeddings, similarity metrics and a vector database to implement semantic search
• ✓Chunk documents into context-preserving segments and blend lexical and semantic scores in hybrid search
• ✓Use an LLM to expand queries and a reranker to re-score retrieved candidates
• ✓Measure retrieval precision, recall and relevance to evaluate and compare your search system
• ✓Combine retrieved context with an LLM for generation, then extend the pipeline to agents and images

• •You implement every retrieval primitive yourself (inverted index, TF-IDF, BM25, hybrid scoring, reranking) rather than calling a RAG framework
• •One continuous project, the Webflyx movie search CLI, grows from plain keyword search into a full RAG pipeline
• •A dedicated evaluation chapter on precision, recall and relevance, a step many RAG tutorials skip
• •Rated 4.7/5 from 145 reviews with 3,879 students enrolled, and runs on free models via OpenRouter so API costs stay near zero
• •Goes past text with a closing multimodal chapter that extends retrieval to images

This page was written from 4 sources, 1 on domains other than boot.dev.

1. 1.boot.dev — learn retrieval augmented generationvendor
2. 2.boot.dev — 766d7ba0 39ed 44e1 993d c0239b4534dbvendor
3. 3.boot.dev — bootdev beat 2025 11vendor
4. 4.scrimba.com — best rag tutorials and courses