Etsy processes vast amounts of unique listing data using machine learning models powered by streaming data from Kafka. Building these streaming applications was a complex and time-consuming manual process for engineers. To streamline this, Etsy developed "Kafka Skills," reusable playbooks for large language models. These skills allow developers and AI agents to collaboratively create new streaming features. The agent generates code, validates configurations, and runs applications locally, significantly reducing manual effort. Observability tools were implemented to track usage and identify areas for improvement. Initially, adoption was low because developers were unaware of the new capabilities and agents didn't proactively offer them. To address this, Etsy improved discoverability by adding more entry points and context-aware routing based on developer vocabulary. Guardrails were also implemented to ensure security and prevent agents from making critical errors by restricting their access and guiding their actions. The goal is to enable faster iteration of machine learning models by simplifying data pipeline development. This ultimately enhances the Etsy user experience by improving search relevance and recommendations.