Migrating to the cloud isn't just about moving servers. For a 150-person professional services company running daily operations through a decade-old system, it meant untangling years of accumulated complexity — a 4TB file store living directly on the application server, scheduled jobs communicating with internal on-premise systems, and a monolithic architecture that was slow, expensive, and fragile.
The goal wasn't just to move. It was to build something that would actually perform.
What We Did
The legacy system ran as a single Linux monolith — application, files, and scheduled jobs all on one large server. That approach had a cost: a monthly AWS bill around $1,500, sluggish response times, and no real path to scale.
We redesigned the architecture from the ground up. The new backend is stateless and distributed, with file storage moved to S3, scheduled jobs offloaded to Lambda and EventBridge, and a highly available RDS setup with Multi-AZ and read replicas. Resources that didn't need to run on the application server simply don't.
One of the more complex challenges was connectivity. Several scheduled jobs needed to communicate with internal on-premise systems that weren't reachable over the public internet. We worked with the client's internal network engineers to configure a site-to-site VPN and routing gateway — ensuring those jobs continued running without any interruption to the business.
CI/CD pipelines and automated vulnerability scanning were established as standard practice throughout.
The Outcome
- Infrastructure costs reduced by over 65% — including two lower environments and a fully highly available RDS with read replication
- Application response times improved from ~900ms to under 300ms at p95
- Zero unplanned downtime during business hours since launch
- A stateless, distributed architecture that scales without the constraints of the old monolith