The legacy platform feeding odds to bookmakers took 13 days to launch a single betting market. I owned the catalogue that took it to minutes.
Sportradar powers live sports data and odds to bookmakers at real scale. The platform underneath had grown fast, and like most systems of that age, prematch and live had evolved as separate stacks: different codebases, different architectures, different logic. For the content teams who launch and extend sports, every new betting market meant leaning on engineering and manually mapping each definition into the model. The average launch took 13 days: engineering time spent on what should be a configuration task, and 13 days a competitor could be live first.
The catalogue is where every market definition lives, and everything downstream depends on its shape. The hardest work wasn't migrating the legacy definitions, it was designing how the new market model should be structured from scratch. Get the schema wrong and you have just repainted the same constraints. I had to design a model general enough to express every market across prematch and live, structured enough to drive automation, and stable enough that consumers and models could build on it without churn.
What shipped was a fully self-serve Markets Catalogue: the legacy definitions migrated in, an API layer that lets consumer systems pull market data directly, and an automation link to model hosting, so the moment a new market is added the relevant sport model maps it and odds calculate with no manual handoff. Content teams configure a market and it flows through to consumers and models on its own. No engineering ticket, no manual mapping.
We built it for one domain, but the real win was the modularity. Designed as a clean general foundation instead of another domain-specific patch, it became something products across the company could build on rather than reinvent. The vision was never a faster catalogue. It was a modular platform that retires the legacy constraints one domain at a time.