Training

Building autonomous AI agents with LangChain and LangGraph

Generative AI has brought many new opportunities and tools to every area of business and IT. But the ready-made tools on the market don’t always help an organization automate its specific tasks or processes effectively and to a high standard. Autonomous agents, equipped with the organization’s internal knowledge and a set of tools that extend their capabilities, are a great fit for this purpose. LangChain and LangGraph make up a mature platform for building flexible autonomous AI agents and scaling their use across an organization. In this training, we’ll explore what the platform can do and learn, through hands-on practice, how to develop and run AI agents of varying architecture and complexity.

Goals and objectives

  • Get to know the capabilities of the LangChain and LangGraph platform for building autonomous AI agents
  • Learn to connect additional knowledge sources and tools to AI agents
  • Develop several AI agents of varying architecture and complexity through hands-on practice
  • Learn to optimize AI agents for cost, quality, and speed
Target audience
Java or Python developers, architects, tech leads
Price
$450per participant
Date and time
8 modules of 3 hours each
Trainer
Mikalai AlimenkouIndependent consultant

Book for your company

Detailed program

Part 1

  • Generative AI and the Transformer architecture
  • How LLMs work and the main stages of their training
  • Capabilities and limitations of LLMs
  • LLMs on the market and how they are classified

Part 2

  • Overview of OpenAI models and their hosting options
  • Creating an account and using the playground
  • Comparison with Anthropic models
  • Deploying models on Amazon Bedrock
  • Prompt engineering basics
  • How to use the OpenAI API and its main capabilities

Part 3

  • Introduction to LangChain
  • Setting up the environment for working with LangChain
  • A quick overview of the main features
  • Using prompt templates
  • Building a simple LangChain chain in chat mode
  • Using structured outputs

Part 4

  • Configuring memory and the different memory modes
  • Implementing persistent memory
  • RAG for adding private knowledge to the context
  • Embeddings and integration with various vector stores
  • Different RAG strategies

Part 5

  • Defining an autonomous agent
  • The ReAct pattern
  • Adding tools
  • Existing tools and how to integrate them
  • The MCP protocol and its advantages
  • Integrating with existing MCP servers
  • Implementing a simple agent with a set of tools

Part 6

  • Limitations of the ReAct pattern
  • Other architectural approaches to building agents
  • Plan-Execute, ReWOO, and multi-agent architectures
  • Introduction to LangGraph for agent orchestration
  • Building a Plan-Execute agent with LangGraph
  • Managing persistent long-term memory

Part 7

  • Using deep agents for long-running tasks
  • Subagents, skills, and sandboxes
  • The agentic harness architecture in Anthropic’s Managed Agents
  • Deploying agents with Managed Agents

Part 8

  • Monitoring agents
  • Testing agent output quality
  • Deploying and scaling agents
  • Optimizing integrations for cost and speed

Past runs

  • June 2-25, 2026Online8 sessions on Zoom on Tuesdays and Thursdays (4:00 to 7:00 p.m., Kyiv time).
  • February 11-25, 2026Online7 sessions on Zoom on Mondays, Wednesdays, and Fridays (4:00 to 7:00 p.m., Kyiv time).
  • November 26 - December 8, 2025Online6 sessions on Zoom on Mondays, Wednesdays, and Fridays (4:00 to 7:00 p.m., Kyiv time).
  • November 3-14, 2025Online6 sessions on Zoom on Mondays, Wednesdays, and Fridays (4:00 to 7:00 p.m., Kyiv time).