How Deutsche Börse built a generative AI tool to tackle the large-scale migration of Zeppelin notebooks to Databricks

Deutsche Börse automated large-scale Zeppelin to Databricks notebook migration using a hybrid approach combining structural conversion with AI-powered logic reconstruction.

• •Deterministic structural conversion handles paragraph format, interpreter syntax, and metadata translation to Databricks standards
• •AI-assisted logic reconstruction uses Claude to rebuild business logic with context about the original Zeppelin environment
• •Migration time reduced from hours to 15-20 minutes per notebook, making the tool accessible to non-specialist business users
• •Hybrid approach of automating deterministic elements and using LLMs for heterogeneous logic avoids brittleness of rule-based systems
• •Context-rich prompts encoding custom interpreters and data sources are critical for producing accurate, usable AI output