🎥 Recorded live at the MLOps World | GenAI Summit 2025 — Austin, TX (October 8, 2025) Session Title: Code-Guided Agents for Legacy System Modernization Speaker: Calvin Smith, Senior Researcher — Agent R&D, OpenHands Talk Track: AI Agents for Developer Productivity Abstract: Modernizing legacy systems is one of the toughest challenges in enterprise software — and AI agents are changing how we approach it. In this talk, Calvin Smith from OpenHands presents a pragmatic approach to legacy code modernization using code-guided AI agents. After early attempts at full-codebase automation led to chaos, Calvin’s team discovered a scalable solution: combining static dependency analysis with intelligent agents that work in incremental, reviewable chunks. The result? Complex refactors are now broken into manageable pull requests — turning a high-risk, all-at-once rewrite into a controlled, transparent transformation process. This session dives into the technical patterns that make it possible for AI systems and human engineers to collaborate effectively, modernizing massive codebases with precision and confidence. What you’ll learn: • How to use static dependency analysis to guide agentic code transformations • How multiple agents can collaborate on complex modernization tasks • How to structure AI-led code refactors into safe, reviewable PRs • The feedback loop between static analysis, AI agents, and human expertise • Why hybrid “AI + human-in-the-loop” systems outperform autonomous ones.