learning-rag

My hands-on learning repo for Retrieval-Augmented Generation (RAG): notes, experiments, and small projects.

Goal

Build a solid, practical understanding of RAG by progressing from core concepts to a working end-to-end pipeline, and be ready to talk about it confidently in interviews.

Learning stages

Each stage below has its own folder with learning goals, a suggested exercise, and a set of related interview questions.

Stage 1: Foundations covers embeddings, vector similarity, and tokenization basics. Stage 2: Vector Stores and Basic Retrieval covers building a first end-to-end RAG pipeline. Stage 3: Chunking and Retrieval Strategies covers chunking choices, hybrid search, and reranking. Stage 4: Evaluation covers measuring retrieval and answer quality, and debugging hallucinations. Stage 5: Agentic RAG covers multi-step agents, tool use, and safety concerns.

Suggested structure

learning-rag/
stage-1-foundations/          learning notes + interview questions
stage-2-vector-stores/        learning notes + interview questions
stage-3-retrieval-strategies/ learning notes + interview questions
stage-4-evaluation/           learning notes + interview questions
stage-5-agentic-rag/          learning notes + interview questions
notebooks/                    exploratory Jupyter notebooks for each concept
src/                          reusable Python modules (chunking, embedding, retrieval, generation)
data/                         sample documents and datasets used for experiments

Progress log

Set up environment and dependencies. Build first basic RAG pipeline. Experiment with different vector stores. Try different chunking strategies. Add evaluation metrics. Build a simple agentic RAG loop.