700 Government Customers, 0 Service Interruptions: A GovTech Cloud Migration
A ground-up target environment plus 12 disciplined migration waves in ~7 weeks, every cluster live, zero customer-facing failures, no schedule slippage.
What made this hard.
- Complete target environment had to be built and validated from scratch.
- ~700 live government customers running public-facing services.
- Nine customer clusters plus shared microservices with dependencies.
- Overnight windows on a twice-weekly cadence across time zones.
- Live risks, authentication dependencies, cert expiries, env sync.
- Concurrent new-platform buildout competing for attention.
How we engineered the solution.
Target environment fully provisioned and validated before any migration.
Phased wave rollout: shared services, then platform, then clusters.
Change Control Board approval per wave with executive steering updates.
Load-testing buffer after the first cluster, with the auth issue engineered out of later waves.
The biggest shift wasn't the migration itself. It was refusing to let the migration carry any risk the team could remove beforehand, the destination existed and was validated before anyone moved.
What this engagement proved.
Large migrations succeed on sequencing and communication.
Building the destination first converts cutover risk into a solved problem.
Strong partners catch the first issue and engineer it out permanently.
For public-facing services, zero downtime is the entry requirement.