Dismantling Datadog’s Monolith with Agentic AI

AI can write code, but how does it do with a large, organically built codebase with years of development? Datadog’s monolithic webapp had significant scale and support challenges, so we knew the time had come to break it up. But when the code has been written, revised and tweaked so many times (as we keep adding new features and capabilities), by so many teams, that becomes much harder.

Nick Isaacs, Senior Software Engineer at Datadog, took on this challenge with Claude, GPT 5 and LangGraph at the ready. He’ll walk through how the team used a constellation of AI tools to decompose the monolith bit by bit, refactoring and rearchitecting the core of Datadog’s oldest codebase. You’ll learn how to spot and mitigate common challenges, how to create helpful patterns for your engineering teams, and come away with expert tips to get more reliable and accurate code. Come to learn how Datadog tackled this tricky problem and leave with insights to save you time and effort on your next major refactoring.