RAG, LangChain & LangGraph

Overview

RAG, LangChain & LangGraph focuses on building retrieval and agentic applications with the LangChain ecosystem. You will work through an ingestion pipeline, a retriever bake-off,…

What you will learn

  • Design a RAG architecture and build a document ingestion pipeline with metadata (Module 1).
  • Configure vector stores and select similarity, MMR, hybrid, and parent-document retrieval by use case (Module 2).
  • Compose chains with LCEL, output parsers, and streaming (Module 3).
  • Build conversational RAG with citations, query rewriting, and re-ranking (Module 4).

Prerequisites

  • Required: Python proficiency and prior LLM API experience. Laptop able to run Python and Docker. API keys and a LangSmith account [confirm whether provided].
  • Helpful: Completion of Generative AI for Developers or equivalent. If you have not called an LLM API before, start there.

Curriculum

  1. Module 1: RAG Fundamentals and Architecture

    • Topics: Knowledge cutoffs and hallucination; RAG vs fine-tuning; system components; loaders for PDF, web, and databases; chunking and overlap; metadata design.
    • Deliverable: Pipeline code. Completion check: Module 1 assessment.

    Hands-on:

    • Practice: Ingestion Pipeline. Load, chunk, and embed a mixed corpus.

  2. Module 2: Embeddings, Vector Stores, and Retrieval

    • Topics: Embedding model selection; vector store options; index management; similarity, MMR, and hybrid search; metadata filtering; parent-document and multi-vector retrieval.
    • Deliverable: Precision comparison. Completion check: Module 2 assessment.

    Hands-on:

    • Practice: Retriever Bake-off. Three strategies measured on a question set.

  3. Module 3: LangChain Core: Models, Prompts, and LCEL

    • Topics: Chat models and prompt templates; output parsers; runnables and LCEL; sequential and parallel chains; streaming and batching; error handling.
    • Deliverable: Chain code. Completion check: Module 3 assessment.

    Hands-on:

    • Practice: LCEL Summarizer. Streaming chain with structured parsing.

  4. Module 4: Building and Optimizing RAG with LangChain

    • Topics: Retriever-to-LLM chains; conversational RAG with history; citations; query rewriting and HyDE; re-ranking; self-querying retrievers.
    • Deliverable: RAG assistant. Completion check: Module 4 assessment.

    Hands-on:

    • Practice: Conversational Knowledge Assistant. History-aware RAG with citations and re-ranking.

  5. Module 5: Tools, Agents, and Function Calling in LangChain

    • Topics: Defining tools and schemas; built-in toolkits; tool error handling; tool-calling agents; ReAct agents; limits of linear agents.
    • Deliverable: Agent code. Completion check: Module 5 assessment.

    Hands-on:

    • Practice: Research Agent. Search, calculate, and retrieve from the vector store.

  6. Module 6: LangGraph Fundamentals

    • Topics: State, nodes, and edges; conditional routing; cycles; StateGraph and reducers; compiling and running; visualizing graphs.
    • Deliverable: Routing graph. Completion check: Module 6 assessment.

    Hands-on:

    • Practice: Customer Query Router. Classify and route to specialized nodes.

  7. Module 7: Advanced LangGraph: Memory, Persistence, and Human-in-the-Loop

    • Topics: Checkpointers and threads; short-term vs long-term memory; time travel and replay; interrupts and approvals; multi-agent and supervisor graphs; subgraphs.
    • Deliverable: Persistent graph. Completion check: Module 7 assessment.

    Hands-on:

    • Practice: Approval-Gated Agent. Multi-agent graph with persistence and an approval step.

  8. Module 8: Evaluation, Tracing, and Deployment

    • Topics: Tracing runs in LangSmith; datasets and evaluators; RAG metrics; serving graphs via API; streaming to a UI; cost, latency, and scaling.
    • Deliverable: Deployed service with traces. Completion check: Module 8 assessment.
    • Capstone (independent work, [CAPSTONE_HOURS] hours): A production RAG agent in LangGraph with multi-step retrieval, tool use, persistent memory, a human approval gate, LangSmith evaluation, and an API deployment.

    Hands-on:

    • Practice: Trace and Deploy. Evaluate on a dataset and deploy behind FastAPI.

Course fee

$699 $399

What's included

  • Live cohort + recordings: Attend every session live and revisit the recordings for one year.
  • Real lab environment: Practise on a hands-on environment that mirrors real projects.
  • Live doubt-clearing support: Ask questions and get answers in real time during every session.
  • Assessments and lab activities: Apply what you learn through graded assessments and guided labs.
  • Learning assets: Keep the slides, notes, and reference material for good.
  • Role-readiness guidance: Mentor guidance to get you ready for the role you are targeting.
  • Verifiable certificate: Earn a shareable certificate on completing the hands-on capstone.
  • Global peer community: Learn with a global cohort and stay connected after the program.

In short

RAG, LangChain & LangGraph is a 24-hour instructor-led course covering Module 1: RAG Fundamentals and Architecture, Module 2: Embeddings, Vector Stores, and Retrieval, Module 3: LangChain Core: Models, Prompts, and LCEL. It includes guided practical labs.

Key facts

  • Course: RAG, LangChain & LangGraph
  • Duration: 24 hours
  • Modules: 8
  • Format: instructor-led live training

Frequently asked questions

Do I need prior LangChain experience?

No, but you need prior LLM API experience in Python. LangChain and LangGraph are taught from the ground up.

Which vector database is used?

Labs run on an open-source store such as Chroma or FAISS. Patterns transfer to hosted options like Pinecone.

How much LangGraph is covered?

Three modules: fundamentals, advanced memory and human-in-the-loop, and deployment. You build and deploy a multi-agent graph.

How is this different from Generative AI for Developers?

That course builds a first RAG app with plain API calls. This course goes deep into advanced retrieval, framework composition, and graph-based agents.

Is LangSmith required?

Module 8 uses LangSmith for tracing and evaluation. [Confirm whether a free tier is sufficient or an account is provided.]

Do I deploy to a cloud provider?

You deploy behind a FastAPI endpoint locally or on a host of your choice. Cloud costs, if any, are yours.

What do I keep?

The full repository: ingestion, retrievers, chains, agents, graphs, evaluation datasets, and deployment code.

Is a corporate version available?

Yes, including retrieval labs on your document types and stack.

Topics covered

RAG, LangChain & LangGraph, Artificial intelligence, Corporate training