Define in code: Build Your Semantic Layer with AI (Live Walkthrough)

Define your semantic layer in code, explained and demoed live. In this session, Johan Baltzar (Co-founder & CEO) and Oscar Albrecht (Senior Software Engineer) walk through why teams are moving to a code-first semantic layer, then build one from scratch in real time, using AI to speed up the process. Oscar builds a workspace live from an empty repo, connects a data source, defines modules, metrics, entities, and join paths in YAML, and shows how an AI agent can generate an entire semantic layer from documentation alone. You'll also see how code-defined and UI-defined metrics can work side by side, how syncs are validated and rejected on errors, and how agents can query and report on the semantic layer once it's built. What's covered: Why code offers audit history, peer review, and reusability that the UI can't The Steep Code YAML format: modules, dimensions, metrics, entities, and join paths Using an AI agent to scaffold an entire semantic layer from scratch How Steep validates syncs and handles errors vs. warnings Mixing code-defined and UI-defined metrics for governed flexibility Live Q&A: repo structure, Open Semantic Interchange support, and migrating in or out of Steep Timestamps: 00:00 Welcome and agenda 01:10 Intro to Steep and the semantic layer 03:19 Steep as Code overview 07:08 Live demo: building a semantic layer from scratch 30:44 Q&A