Deutsche Börse builds Databricks App to migrate 2,000+ notebooks with AI
Converts hours of manual work into 15–20 minute AI-assisted rewrites.
Deutsche Börse Group faced a massive notebook migration deadline: Cloudera is decommissioning Zeppelin by 2027, affecting 2,000+ users relying on complex notebooks for clearing and trading data analytics. Manual rewriting would take years, so the StatistiX team built a Databricks App that separates structural conversion from logic reconstruction. The converter maps Zeppelin paragraphs to Databricks cells, translates interpreter syntax, and reformats metadata into valid .ipynb JSON — all without altering original SQL, Python, widgets, or Oracle/HDFS references.
For each notebook, the app automatically generates a context-aware prompt for Databricks Genie, which then reconstructs the business logic interactively. Users simply export, upload, convert, and launch Genie. This approach reduced redevelopment time from hours to 15–20 minutes per notebook. The app uses a shadcn UI frontend and runs on Databricks Apps, avoiding separate infrastructure. By intentionally not automating logic rewrites, the team preserved institutional knowledge and avoided introducing errors.
- Handles 2,000+ notebooks with a 2027 deadline, reducing rewrite time from hours to 15–20 minutes per notebook.
- Separates structural conversion (deterministic, automated) from logic reconstruction (handled by Databricks Genie with context-aware prompts).
- Preserves original SQL, Python, visualizations, widgets, and Oracle/HDFS references to avoid errors and maintain trust.
Why It Matters
Practical, large-scale AI-assisted migration saves months of developer time and ensures business continuity.