Advanced RAG: Vector Databases, Hybrid Search, GraphRAG & Agentic RAG
Coming soon BeginnerIntermediateAdvanced
Get notified on YouTube Udemy certificate: coming soon
- Udemy
- Coming soon
- Lessons
- 45
- Planned length
- About 12 h 8 min
- Projects
- 3
- Practicals
- Planned: a hands-on practical and a Udemy assignment in every section, plus a capstone
What you will be able to do
- Load a folder of mixed documents
- Pick a chunking strategy with a quick test
- Parse tables and scanned pages correctly
- Filter retrieval by source, date and section
- Measure which strategy wins on your data
- Embed sentences on your laptop and compare them
Syllabus
RAG: Retrieval-Augmented Generation: Ingestion and Chunking
- Loading PDFs, Docs and Web Pages Load a folder of mixed documents
- Chunking Basics: Fixed, Recursive and Structure-Aware Pick a chunking strategy with a quick test
- Hard PDFs: Tables, Scans and OCR Parse tables and scanned pages correctly
- Metadata, Filters and Self-Querying Filter retrieval by source, date and section
- Semantic and Advanced Chunking: Parent-Document, Contextual, Late Measure which strategy wins on your data
RAG: Retrieval-Augmented Generation: Embeddings
- Embeddings Explained With Code, Free and Local Embed sentences on your laptop and compare them
- Choosing an Embedding Model: Paid vs Open, Multilingual, ColBERT Pick a model by quality, language and cost
- Fine-Tune an Embedding Model on Your Domain Lift retrieval accuracy with domain pairs
RAG: Retrieval-Augmented Generation: Vector Databases (One Lesson per Tool)
- What a Vector Database Does: ANN, HNSW, IVF Explain indexes and the speed/recall trade-off
- Chroma Tutorial Store and query embeddings locally
- FAISS Tutorial Build an in-memory index for millions of vectors
- pgvector: Vectors in Postgres Add semantic search to an existing Postgres app
- Pinecone Tutorial Run a managed serverless index
- Qdrant Tutorial Use payload filters and hybrid search in Qdrant
- Weaviate Tutorial Use built-in vectorizers and hybrid search
- Milvus Tutorial Run a distributed vector DB for large collections
- Elasticsearch and OpenSearch for Vector and Hybrid Search Add vectors to an existing search stack
- Choose a Vector DB (incl. LanceDB, Redis, Atlas): Benchmark on Your Data Pick a DB from measured recall, latency and cost
RAG: Retrieval-Augmented Generation: Retrieval Types
- Dense Retrieval Retrieve by meaning with embeddings
- Sparse Retrieval and BM25 Catch exact terms, codes and names
- Hybrid Search Combine dense and BM25 with rank fusion
- Rerankers: Cross-Encoders, Cohere Rerank, BGE Re-score top results for precision
- Query Rewriting and Diversity: Multi-Query, HyDE, MMR Fix vague questions and stop duplicate chunks
RAG: Retrieval-Augmented Generation: RAG Types
- Naive RAG Build the baseline everything else improves on
- Advanced RAG: Rewriting, Reranking and Chat History Upgrade naive RAG and handle follow-up questions
- Modular RAG Assemble RAG from swappable modules
- Agentic RAG Let an agent decide what and when to retrieve
- Graph RAG: GraphRAG, LightRAG and Neo4j Answer whole-corpus questions with a knowledge graph
- Multimodal RAG: Images, Tables and Slides Retrieve from charts and images, not just text
- Corrective RAG (CRAG) Detect bad retrieval and fall back to web search
- Self-RAG Make the model critique its own retrieval and answer
- Adaptive RAG: Route by Question Type Route simple, complex and off-topic questions differently
RAG: Retrieval-Augmented Generation: RAG Frameworks
- No-Code RAG: NotebookLM, Dify, Flowise Build a document assistant without code
- Haystack Tutorial Build a RAG pipeline in Haystack
- LangChain vs LlamaIndex vs Haystack vs Plain Python Choose a framework for your team
RAG: Retrieval-Augmented Generation: RAG Evaluation
- Why RAG Needs Evaluation Spot the difference between "looks right" and "is right"
- RAG Metrics: Faithfulness, Relevance, Recall@k Pick the metric for each failure
RAG: Retrieval-Augmented Generation: RAG in Production
- From Notebook to Product: RAG Checklist List what changes between a demo and a product
- RAG Security: Permissions, PII, Poisoned Documents Enforce who can see which document
- Monitor and Trace RAG in Production (Phoenix, TruLens, Langfuse) Trace retrievals and alert on falling answer quality
- Scaling RAG: Fresh Indexes, Latency, Cost, Millions of Docs Plan incremental indexing, caching and sharding
- Capstone: Company Docs Support Bot Ship a cited, evaluated support bot over company docs
Frontier and Advanced: Reasoning Models and Inference
- Long Context vs RAG: Measure Cost and Accuracy Measure when to stuff the context and when to retrieve
Frontier and Advanced: Frontier Agent Engineering
- Knowledge Graphs for Agents: Query and Update Tools Give an agent a graph it can query and update
- Project: Agent With Knowledge-Graph Memory Ship an agent that builds and queries its own graph
Take before
Lesson list from the ADITEJUS curriculum. Planned lengths are estimates per lesson; Udemy shows the final length.