From Alerts to Agents: Rethinking Data Quality with Claude Code

Data quality failures are an unavoidable fact of life for data engineers, but the time spent debugging them doesn't have to be. When a pipeline breaks, the root cause is rarely obvious. It could be a schema change upstream, a missing test, or a subtle code change buried somewhere in your DAG. In this live demo, we show how Bauplan and Claude Code turn that painful manual process into a fully automated and future proof workflow. Bauplan gives you a Python-native lakehouse platform with Git-for-Data branching, so your AI agent can investigate, fix, and test in complete isolation, without ever touching production tables. Claude Code handles the rest: reading logs, identifying the root cause, applying a minimal fix, and writing data quality tests so the problem never comes back. We walk through a real pipeline failure end to end. From an upstream schema change that breaks a downstream aggregation, to a fully automated diagnosis, fix, and future-proofing workflow. No manual debugging, no production risk, and a complete audit trail in both code and data. 0:00 Introduction & Welcome 1:33 The Two Types of Data Quality Failures 4:01 Demo Pipeline Overview & Infrastructure Requirements 8:52 Live Demo: From Failing Pipeline to Root Cause 20:06 Live Demo: Fixing, Testing & Merging to Production 24:36 Q&A 39:30 Closing Remarks & Next Webinar Announcement