Building Agentic AI with LangGraph

Session Abstract
While standard RAG pipelines and linear chains have dominated the early GenAI era, the industry is shifting toward Agentic Workflows—systems that can reason, loop, and correct themselves. However, building these systems with traditional tools often leads to "black-box" behavior and unreliability.

In this session, we dive into LangGraph, the low-level orchestration framework designed to bring structure to agentic chaos. We will explore how to move beyond basic AgentExecutor patterns to build custom, stateful, and multi-agent systems. You’ll learn how to model AI logic as a StateGraph, implement robust Human-in-the-Loop checkpoints, and manage long-term Memory for complex, multi-turn interactions.

What We’ll Cover
-The Blueprint of Agency: Understanding Nodes (computation), Edges (control flow), and State (the shared memory).
-Breaking the Chain: Why cyclic graphs are superior to linear pipelines for reasoning and self-correction.
-Human-in-the-Loop: Designing "interrupts" and "time travel" to allow manual oversight and state editing in production.
-Multi-Agent Architectures: Orchestrating a "team" of specialized agents (e.g., Researcher, Coder, Reviewer) using hierarchical and sequential patterns.
-Persistence & Scalability: Leveraging checkpoints to ensure agents can resume long-running tasks after failures.

LangGraph

Discover how to build the next generation of AI agents using LangGraph. While standard chains are great for chatbots, LangGraph unlocks the ability to create loops and handle state—essential for agents that actually "do" things.

I will showcase a live breakdown of an Agentic workflow designed for the insurance industry, demonstrating how to coordinate multiple AI agents to solve a single complex problem efficiently.

Building Agentic AI with LangGraph

Session Abstract
While standard RAG pipelines and linear chains have dominated the early GenAI era, the industry is shifting toward Agentic Workflows—systems that can reason, loop, and correct themselves. However, building these systems with traditional tools often leads to "black-box" behavior and unreliability.

In this session, I will dive into LangGraph, the low-level orchestration framework designed to bring structure to agentic chaos. We will explore how to move beyond basic AgentExecutor patterns to build custom, stateful, and multi-agent systems. You’ll learn how to model AI logic as a StateGraph, implement robust Human-in-the-Loop checkpoints, and manage long-term Memory for complex, multi-turn interactions.

What I will Cover
-The Blueprint of Agency: Understanding Nodes (computation), Edges (control flow), and State (the shared memory).
-Breaking the Chain: Why cyclic graphs are superior to linear pipelines for reasoning and self-correction.
-Human-in-the-Loop: Designing "interrupts" and "time travel" to allow manual oversight and state editing in production.
-Multi-Agent Architectures: Orchestrating a "team" of specialized agents (e.g., Researcher, Coder, Reviewer) using hierarchical and sequential patterns.
-Persistence & Scalability: Leveraging checkpoints to ensure agents can resume long-running tasks after failures.

Data Leadership World Summit 1.0

Building Agentic AI with LangGraph

Session Abstract
While standard RAG pipelines and linear chains have dominated the early GenAI era, the industry is shifting toward Agentic Workflows—systems that can reason, loop, and correct themselves. However, building these systems with traditional tools often leads to "black-box" behavior and unreliability.

In this session, we dive into LangGraph, the low-level orchestration framework designed to bring structure to agentic chaos. We will explore how to move beyond basic AgentExecutor patterns to build custom, stateful, and multi-agent systems. You’ll learn how to model AI logic as a StateGraph, implement robust Human-in-the-Loop checkpoints, and manage long-term Memory for complex, multi-turn interactions.

What We’ll Cover
-The Blueprint of Agency: Understanding Nodes (computation), Edges (control flow), and State (the shared memory).
-Breaking the Chain: Why cyclic graphs are superior to linear pipelines for reasoning and self-correction.
-Human-in-the-Loop: Designing "interrupts" and "time travel" to allow manual oversight and state editing in production.
-Multi-Agent Architectures: Orchestrating a "team" of specialized agents (e.g., Researcher, Coder, Reviewer) using hierarchical and sequential patterns.
-Persistence & Scalability: Leveraging checkpoints to ensure agents can resume long-running tasks after failures.