Agentic AI
Build Agents that Think, Decide, and Act. Master LangGraph and turn LLMs into Autonomous Teammates!
13 topics
50 hours
Intermediate
Certificate
The course
Master the art of creating production-ready agentic AI systems using LangGraph, the leading framework for building stateful, multi-step, decision-making agents powered by LLMs.
This hands-on, practical course takes you step-by-step from basic graphs to sophisticated multi-agent teams — with real code, mini-projects, and deployment at every stage.
Start with simple stateful agents and chatbots, then add tool calling, conditional routing, and dynamic decisions. Progress to iterative loops, persistent memory, self-correction, and human-in-the-loop controls. Build personalized assistants, multi-agent teams with supervisors and workers, classic RAG for document grounding, and finally intelligent agentic RAG — where agents autonomously decide when/how to retrieve, grade results, rewrite queries, self-correct, fall back to web search, and reflect — before deploying production-grade systems.
Pre requisites : Proficient in Python programming language, basic layman interaction with any of the LLMs out there (ChatGpt, Grok, Gemini) and lots of optimism!
The journey
- What is an LLM? (next-word prediction basics)
- Prompt vs. Completion
- Chat format: System / User / Assistant roles
- Build: Simple role-based chat + continuation
- Why messages matter for agents
- Install langchain-core + LLM wrapper
- Chat models & message invocation
- Messages types & conversation history
- Tools & bind_tools
- ChatPromptTemplate + MessagesPlaceholder
- LCEL (pipes) patterns
- Build: An agent loop manually using what we have learnt in LangChain
- What is Agentic AI? Autonomous agents, goal-oriented systems
- LangGraph vs LangChain (LCEL limitations in cycles/branches)
- Graphs, nodes, edges, state
- Build: Hello World agent + mini chatbot project
- Python env, LangGraph + langchain-core + LLM provider
- API keys, Jupyter/VS Code setup
- Build: Verification script + reusable project template
- Graph API basics, nodes/edges
- State: MessagesState vs custom TypedDict
- Compilation, invocation, streaming
- Build: Linear Q&A agent with mock tool
- Routing based on LLM output/tool calls
- Error handling, fallbacks
- Multi-tool router
- Build: Weather/cache decision agent
- Loops for iteration/self-correction
- Checkpoints (MemorySaver), resuming
- Max iterations, loop control
- Refine-answer-until-good loop
- Build: Iterative research agent
- Breakpoints, interrupts, user input
- Streaming intermediate outputs
- Approval before action
- Build: Email drafter with confirmation
- Custom nodes, logging/metrics
- Advanced state schemas + reducers
- Tool binding/custom tools
- Personalised memory agent
- Build: Step-logging assistant
- Supervisor-worker, hierarchical/parallel patterns
- Message passing, shared state
- Researcher + summariser team
- Build: Trip planner (planner/booker/reviewer)
- LangGraph Cloud, FastAPI/AWS/local
- Monitoring (LangSmith), security, ethics
- FastAPI endpoint for agent
- Build: Deployed support bot
- What is RAG? Why needed (hallucinations, private/current data)
- Core steps: Load → Split → Embed → Index → Retrieve → Augment → Generate
- LangChain LCEL chain for naive RAG
- Vector stores (Chroma/FAISS), embeddings, chunking
- Irrelevant retrieval, context loss
- Build & test full RAG chain on PDFs/docs
- Build: Q&A bot over your documents + simple UI
- Limitations of linear RAG
- Agentic vs naive RAG: Decision-making, self-correction, adaptive/corrective patterns
- Retriever as tool + grading/rewriting/fallback nodes
- Conditional edges (retrieve? grade? rewrite? web fallback?)
- Cycles for reflection/iteration
- Basic agentic RAG graph → add grading → add query rewrite
- Build: Self-healing knowledge base agent (local docs + web fallback)
The skills
AgenticAI
LangGraph
AgenticRAG
HumanInTheLoop
AIAutomation
MultiAgentSystems
LangChain
AIOrchestration
The projects
Project 1Travel Concierge Agent
Build a smart, multi-agent travel planning system using LangGraph that acts as a personal concierge. The agent team takes user inputs, researches real-time information via tools, collaborates to create a complete itinerary, refines plans iteratively, seeks human approval on major decisions, remembers user preferences across sessions, and delivers a polished, actionable travel plan.