22-25 April 2026 
Optimize Delivery, Not Just Queries: Model Driven Data Engineering in Practice
Proposed session for SQLBits 2026TL; DR
Model-driven delivery reduces spec drift by capturing requirements in a reviewable data model and generating SQL, pipelines, tests, and docs from approved templates. Result: faster releases, lower cost, and more reliable data platforms across any stack.
Session Details
Optimizing a data platform is usually less about tools and more about how you deliver changes safely and repeatedly across SQL Server, Fabric, Databricks, Snowflake, or cloud-native stacks on Azure and AWS. One of the biggest hidden costs is “spec drift”: requirements live in text, implementations live in SQL or Python, and over time they fall out of sync. Reviews become slow, testing becomes fragile, and releases get riskier.
This session presents a practical, model driven approach that improves speed, cost, and reliability without forcing you into a new technology. You capture specifications as a visual data model enriched with metadata that business and engineering can review quickly. In parallel, you standardize query and pipeline patterns as templates. The platform-specific SQL, transformations, tests, and documentation are then generated from the approved model plus approved templates, giving you predictable output and consistent performance patterns.
We will show an existing implementation and how to convert hand-written SQL into reusable templates (including how AI can help extract patterns). By storing examples in the model and generating test scenarios from them, you keep logic aligned with expectations and prevent regressions as you scale.
You will leave with a repeatable method to keep specs, code, and tests aligned, so you can ship faster, test earlier, and release with confidence on the platform you already use.
This session presents a practical, model driven approach that improves speed, cost, and reliability without forcing you into a new technology. You capture specifications as a visual data model enriched with metadata that business and engineering can review quickly. In parallel, you standardize query and pipeline patterns as templates. The platform-specific SQL, transformations, tests, and documentation are then generated from the approved model plus approved templates, giving you predictable output and consistent performance patterns.
We will show an existing implementation and how to convert hand-written SQL into reusable templates (including how AI can help extract patterns). By storing examples in the model and generating test scenarios from them, you keep logic aligned with expectations and prevent regressions as you scale.
You will leave with a repeatable method to keep specs, code, and tests aligned, so you can ship faster, test earlier, and release with confidence on the platform you already use.
3 things you'll get out of this session
A model-first workflow that keeps specs, code, and tests aligned
Reusable templates to generate consistent SQL, pipelines, and documentation
Faster, safer releases with less rework on your current platform