Roadmap
The course mirrors the Python for Everyone architecture: chapters grouped into modules, each chapter ending with exercises, practice, interview prep, a project increment, and a GenAI thought-process journal.
Module 1: RAG Foundations
Why RAG exists, the query lifecycle, RAG architecture families, and how to read system diagrams.
Outcome: explain a RAG system end to end.
Module 2: Documents and Ingestion
Source documents, parsing, cleaning, metadata, and repeatable ingestion pipelines.
Outcome: turn messy source material into traceable knowledge units.
Module 3: Chunking
Fixed, recursive, semantic, and hierarchical chunking, plus failure analysis.
Outcome: choose chunking strategies based on retrieval behavior.
Module 4: Embeddings and Similarity
Vectors, cosine similarity, local TF-IDF + SVD, neural embeddings, drift, and reindexing.
Outcome: reason about embedding choices and limitations.
Module 5: Vector Stores and Retrieval
ChromaDB, top-k retrieval, hybrid search, BM25, reranking, and retrieval trade-offs.
Outcome: design and debug retrieval stacks.
Module 6: Prompting for RAG
Context injection, grounded system prompts, citations, refusal behavior, and prompt testing.
Outcome: write prompts that are grounded and testable.
Module 7: Evaluation
Recall@K, MRR, faithfulness, groundedness, answer relevance, and failure analysis.
Outcome: measure RAG behavior instead of demoing by vibes.
Module 8: Production RAG
Access control, auditability, observability, cost, deployment patterns, and capstone design.
Outcome: defend a production RAG architecture and portfolio project.
Every Chapter Includes
Lesson, quiz, exercises, practice, interview prep, project, and GenAI thought-process journal.
Outcome: learn like a responsible new-age AI builder.
Chapter Path
Why RAG Exists
Hallucinations, private knowledge, trusted context, and citations.
The RAG Query Lifecycle
Question, retrieval, context, prompt, answer, evaluation.
Types of RAG
Naive, advanced, agentic, graph, corrective, adaptive.
RAG Execution Pipelines
Offline indexing, online query execution, gates, traces, and production operations.
Chapter 5Source Documents
Source formats, governance, access, freshness, and synthetic corpora.
Chapter 6Parsing and Cleaning
Parser contracts, cleaning rules, quality gates, and regression tests.
Chapter 7Metadata Design
Source, section, role, access, version, timestamp, and retrieval controls.
Chapter 8Ingestion and Indexing Pipelines
Batch ingestion, incremental updates, re-indexing, logs, rollback, and release gates.
Chapter 9Fixed and Recursive Chunking
Token size, overlap, separators, recursive splitting, and chunk-quality trade-offs.
Chapter 10Semantic and Hierarchical Chunking
Topic-aware boundaries, parent-child retrieval, hierarchy metadata, and context expansion.
Chapter 11Chunking Failure Analysis
Trace inspection, lost context, overlap noise, stale chunks, eval gates, and controlled fixes.
Chapter 12What Embeddings Are
Text-to-vector intuition, similarity, dense vs sparse retrieval, and production embedding controls.
Chapter 13Local TF-IDF + SVD Embeddings
Lexical weighting, latent-topic compression, exact-term strengths, paraphrase limits, and hybrid positioning.
Chapter 14Neural Embeddings
Model choice, provider trade-offs, evals, dimensions, privacy, cost, re-indexing, and rollout planning.
Chapter 15Vector Stores and Indexing Strategies
Collections, ids, metadata, indexes, rebuilds, blue-green release, and rollback.
Chapter 16Retrieval Strategies and Top-K
Top-k tuning, filters, scores, query rewriting, multi-query retrieval, dedupe, and traces.
Chapter 17Hybrid Search and Query Routing
Dense, sparse, BM25, RRF, exact-term protection, routing, and route traces.
Chapter 18Reranking and Retrieval Optimization
Retrieve many, return few, cross-encoder intuition, candidate pools, latency, and eval loops.
Chapter 19Context Injection
Context packs, evidence ordering, compression, prompt boundaries, citations, and token budgets.
Chapter 20Grounded System Prompts
Evidence contracts, refusal behavior, citations, conflicts, templates, prompt evals, and rollout controls.
Chapter 21Prompt Testing
Golden questions, prompt assertions, citation support, refusals, adversarial cases, metrics, and release gates.
Chapter 22Retrieval Metrics
Gold evidence, Recall@K, Precision@K, MRR, nDCG, dashboards, and release gates.
Chapter 23Answer Quality Metrics
Faithfulness, groundedness, relevance, citation correctness, human review, and quality gates.
Chapter 24Failure Analysis
Root-cause analysis, traces, failure taxonomy, owners, fixes, and replay tests.
Chapter 25Access Control and Safety
RBAC retrieval, metadata filters, sensitive data, prompt-injection safety, and audit logs.
Chapter 26Observability and Cost
Traces, latency, token cost, caching, routing, dashboards, and optimization loops.
Chapter 27Deployment Patterns
Static demos, CLI, API services, hosted frontends, vector DB choices, and operations.
Chapter 28End-to-End Production RAG Capstone
Three product builds, five capstone tracks, portfolio packaging, and production trade-offs.
Canonical Source
The editable roadmap lives in
CURRICULUM_MAP.md. Keep this page and
assets/chapters-data.js aligned as chapters are built.