Dmitry Sorokin: Turning ML and Data Pipelines into AI Agents with Kedro and LangGraph

Turning ML and Data Pipelines into AI Agents with Kedro and LangGraph Dmitry Sorokin In this talk, I’ll explore how modern AI agent frameworks like LangGraph and large language models can transform traditional Kedro ML and data pipelines into intelligent, interactive agents. Imagine a pipeline that not only processes data but can decide when to run, handle failures autonomously, and even accept instructions in natural language. We’ll discuss how to wrap Kedro pipelines as LangGraph nodes, using LLMs to orchestrate decisions and manage workflows. I’ll share real-world patterns for building "agentic pipelines" that blend structured, production-grade Kedro engineering with flexible, conversational AI capabilities. The session will cover: - Why treating pipelines as agents is valuable for ML operations and data science teams - Practical architecture: integrating Kedro, LangGraph, and OpenAI models - Demo of a Kedro pipeline controlled via a conversational interface - Lessons learned from building AI-powered pipeline management tools Whether you're working in ML engineering, data science, or AI product development, this talk will give you tools and ideas to modernise your pipelines using Python's latest AI stack. Talk room 2 (200 people) Sat 3:15 pm - 3:45 pm