Speaker
Abstract
We don't need LLMs to write new code. We need them to clean up the mess we already made.
In mature organizations, we have to maintain and migrate the existing codebase. Engineers are constantly balancing new feature development with endless software upkeep.
But what if you could rewrite your codebase, every single day, across thousands of repositories? What if your engineers didn't have to maintain their code?
At Spotify, we are seeing early success using LLMs to perform predictable, repeatable and effortless code migrations.
In this talk, we’ll share how we created an Agentic Migrator that has gotten over 3000 PRs merged across several engineering disciplines. We will tell you how we reason about solving the complexity of LLMs maintaining code at scale. From managing build feedback loops across thousands of repos, to evaluating prompt effectiveness and mastering the sheer complexity of our diverse codebase.
Topics
QCon London 2026 is a three day conference for senior software engineers, architects and team leads. An international program committee of working engineers selects every session. Patterns and practices, not products and pitches.
Part of the track
AI Engineering Hosted by Hien Luu Sr. Engineering Manager @Zoox & Author of MLOps with Ray, Speaker and Conference Committee ChairFrom the same track
Tuesday 17 March
10:35 Whittle (3rd Fl.) Session AI/LLM Reliable Retrieval for Production AI Systems Lan Chu AI Tech Lead and Senior Data Scientist Search is central to many AI systems. Everyone is building RAG and agents right now, but few are building reliable retrieval systems. 11:45 Fleming (3rd Fl.) Session AI Rewriting All of Spotify's Code Base, All the Time Jo Kelly-Fenton, Aleksandar Mitic We don't need LLMs to write new code. We need them to clean up the mess we already made. In mature organizations, we have to maintain and migrate the existing codebase. Engineers are constantly balancing new feature development with endless software upkeep. 13:35 Churchill (Ground Fl.) Session AI/ML Refreshing Stale Code Intelligence Jeff Smith CEO & Co-Founder @Neoteny AI, AI Engineer, Researcher, Author, Ex-Meta/FAIR Coding models are helping software developers move faster than ever, but weirdly, the models themselves are not keeping up. They are trained on months-old snapshots of open source code. They have never seen your internal codebase, let alone the code you wrote yesterday. 14:45 Churchill (Ground Fl.) Session AI Beyond Context Windows: Building Cognitive Memory for AI Agents Karthik Ramgopal Distinguished Engineer & Tech Lead of the Product Engineering Team @LinkedIn, 15+ Years of Experience in Full-Stack Software Development AI agents are rapidly changing how users interact with software, yet most agentic systems today operate with little to no intelligent memory, relying instead on brittle context-window heuristics or short-term state. 15:55 Fleming (3rd Fl.) Session applied ai Building an AI Gateway Without Frameworks: One Platform, Many Agents Amit Navindgi, Jatin Aneja Early AI integrations often start small: wrap an inference API, add a prompt, ship a feature. At Zoox, that approach grew into Cortex, a production AI gateway supporting multiple model providers, multiple modalities, and agentic workflows with dozens of tools, serving over 100 internal clients. 17:05 Whittle (3rd Fl.) Session Async Agents in Production: Failure Modes and Fixes Meryem Arik Co-Founder and CEO @Doubleword (Previously TitanML), Recognized as a Technology Leader in Forbes 30 Under 30, Recovering Physicist As models improve, we are starting to build long-running, asynchronous agents such as deep research agents and browser agents that can execute multi-step workflows autonomously. These systems unlock new use cases, but they fail in ways that short-lived agents do not.