Deutsche Börse Uses AI to Speed Up Cloud Migration

Deutsche Börse runs one of the world's largest exchange groups, covering trading, clearing, and settlement across European markets. Internally, its StatistiX platform serves roughly 2,000 users who rely on it for clearing and trading data every day. For years, that meant working with Zeppelin notebooks on Cloudera, a combination that was once standard across large enterprises that built out big data infrastructure in the 2010s. The problem is that era is ending. Cloudera has deprecated Zeppelin across its platform, with full decommissioning set for 2027. That is not a distant deadline: organizations with thousands of notebooks need to start moving now, because the work is bigger than it looks from the outside. Notebooks in a production analytics environment are not simple files. They contain SQL queries, Python scripts, custom connection logic, scheduling rules, and visualization code, all wired together in ways that reflect years of decisions made by people who may no longer be at the company. Each one is a piece of institutional knowledge, not a commodity file you can auto-convert with a script. Deutsche Börse's team recognized that a rule-based conversion tool would break constantly. Business logic is too varied and too organization-specific for fixed rules to handle reliably. Their solution was to separate the two kinds of work. The mechanical part, converting the file format from Zeppelin's structure to the format Databricks uses, is entirely predictable and can be done automatically. The substantive part, reconstructing the actual logic so it runs correctly in the new environment, requires judgment. For the second part, they used Genie, Databricks' AI assistant that can read data context, understand organizational data structures, and write or rewrite code through conversation. The tool automatically generates a detailed prompt for each notebook, one that includes Deutsche Börse's specific interpreter configurations, data sources, and environment details, and feeds that to Genie to do the reconstruction. Business users upload a notebook, get a prompt, and let Genie do the work. No deep Databricks expertise required. The result is a migration time of 15 to 20 minutes per notebook, down from hours of manual effort. At the scale Deutsche Börse is operating, that difference is enormous: it is the gap between a multi-year engineering project and something a normal team can complete in a reasonable timeframe. Genie itself is worth understanding separately. Databricks launched it in mid-2024 as a conversational analytics tool that lets users ask questions about data in plain language and get answers back as SQL queries, charts, or summaries. It has since expanded into Genie Code, a more autonomous agent that can plan and execute multi-step data work, debug pipelines, and maintain production systems. The Deutsche Börse migration is an early example of using this kind of AI not for new analytical work, but for the unglamorous job of moving existing work from one platform to another. The Cloudera situation is not unique to Deutsche Börse. Cloudera has been steadily phasing out legacy components as it moves customers toward cloud-native architectures, and many organizations that built analytics stacks on Hadoop-era infrastructure are now facing the same forced migration. The scale of the problem tends to be underestimated because infrastructure migrations get attention while the software layer built on top of the infrastructure does not. The practical implication is straightforward. If your organization runs analytics on aging on-premises platforms, and many do, the notebooks, scripts, and queries built on top of those platforms represent a significant hidden migration cost. AI tools have reached the point where they can take on most of the mechanical conversion work. The question is whether teams recognize the problem before it becomes urgent.