We needed to migrate our data pipeline to Azure Event Hubs, which required buy-in from the platform team and the security team. I set up a series of syncs, shared a detailed design doc, and tracked dependencies in a shared spreadsheet. I unblocked most of the hard problems myself — figured out the partitioning strategy, resolved the auth configuration issues — then handed clear tasks to each team. We shipped on time with zero production incidents. The platform team said the integration was cleaner than they expected.
Our search ranking pipeline needed a real-time scoring layer — it touched the ML platform team, the data engineering team, and security. Instead of designing it alone, I ran joint design sessions where each team owned a section of the doc and had veto rights on their domain. When the ML team lacked experience with Event Hubs partitioning, I ran a working session with them — not to solve it for them, but so they could own the solution. We shipped in eleven weeks with forty percent lower p99 latency. Both teams used the partitioning pattern on two subsequent projects without my involvement.