UberConfiguration validation

A global mobility platform operating in 15,000 cities with over 8 million earners, managing a complex architecture of 3,000 microservices with unique regulatory rules.

Configuration sprawl across isolated systems caused costly conflicts, such as drivers passing onboarding checks but failing dispatch rules for specific markets. SQL-based validation queries became unmanageable mazes of recursive joins, while document databases failed to map the complex interdependencies between business rules.

The engineering team built a Knowledge Graph on Neo4j AuraDB that models regulatory and product rules as traversable paths rather than isolated data points. They exposed this graph via the Model Context Protocol, allowing both developers and LLM agents to query dependencies and validate changes across domains in milliseconds. The architecture includes vector indices for semantic search and relies on ACID compliance to ensure validation logic runs in isolated transactions.