Lumen Ridge is a fictional betting operator with an online casino, a sportsbook and a chain of retail betting shops. Before wiring up its real sources, the data team rehearsed the whole platform end to end on synthetic data, with two AI engineers, Claude and Codex, building and reviewing each other's work in public pull requests.
The brief was blunt: seven sources (five APIs, a PostgreSQL database and a file feed), finance-grade reconciliation, and a budget of USD 200 a month with a kill switch at USD 190. Nobody wanted to discover on the real data that a pipeline double counts a bet, that a Power BI refresh takes twenty minutes, or that an idle network gateway quietly costs more than the compute.
So the team built a dress rehearsal. Two generators produced a dirty, realistic month of online play and retail shop activity, complete with late arrivals, re-sent files, conflicting versions and missing settlements. Everything else was real: Azure Databricks with Unity Catalog, Terraform, a budget with alerts, dbt, Power BI, a ClickHouse copy for comparison, and a GitHub repository where every change was argued over until both reviewers wrote AGREE.
The ebook follows that rehearsal chapter by chapter, from the first landed file to the scheduled teardown, with the figures, the costs and the mistakes left in.