Data pipelines ship faster than teams can document, test, or trace them. A column rename quietly breaks three dashboards. A schema drift silently poisons a semantic model. Sound familiar?

This session puts AI to work on the problems data engineers actually lose sleep over - demonstrated live inside Microsoft Fabric with real Lakehouse pipelines, notebooks, and OneLake artifacts:

* Code generation: Convert requirements into Fabric notebooks and pipeline JSON using natural-language prompts. See what works, what needs guardrails, and where human review is non-negotiable.
* Living documentation: Auto-generate and continuously sync technical documentation from actual pipeline and notebook code - so docs never again drift from reality.
* Proactive data quality: Let AI inspect Delta table schemas and data distributions to propose validation rules, detect anomalies, and generate test cases - catching drift *before* it hits production.
* Lineage & impact analysis: Answer "What breaks if I change this column?" by tracing column-level lineage across Fabric pipelines, Spark notebooks, OneLake tables, semantic models, and Power BI reports - and visualize the blast radius before the change ships.
* See Project Osmos (the Fabric data engineering agent) in action. How is this one different to other coding agents?

What you'll take home:

* A practical toolkit: which AI capabilities (inside and outside Microsoft Fabric) are production-ready today and which ones to watch.
* Repeatable patterns for AI-maintained documentation and automated data-quality checks.
* An honest evaluation framework: where AI accelerates data engineering versus where it introduces new risks (and more work).