Project overview
A regulated European fintech was bottlenecked by a monolithic Java application running on legacy VMs. Releases took the better part of a working day, change windows were tightly controlled, and rollbacks were manual and risky.
Challenges
- 45-minute deploys gating every change, with frequent failed cutovers
- No safe progressive delivery — every release was all-or-nothing
- Limited observability across the JVM, infra and customer journeys
- Strict regulatory windows for production changes
Our approach
We re-platformed the workload onto Amazon EKS with a GitOps-driven CI/CD pipeline, progressive delivery, and full-stack observability — all without missing a compliance window.
Containerization & EKS landing zone
- Containerized the Java monolith and split out the highest-churn modules
- Provisioned multi-AZ EKS clusters with Terraform across non-prod and prod accounts
- Hardened IAM, network and pod security baselines from day one
Progressive delivery & GitOps
- Replaced Jenkins jobs with GitHub Actions and Argo CD GitOps sync
- Introduced canary releases via Argo Rollouts with automated metric checks
- Built one-click rollback to the last known-good revision
Observability
- Instrumented the JVM and key flows with OpenTelemetry
- Centralized metrics, logs and traces in Datadog with SLO dashboards
- Wired pager rotations and runbooks into incident response
Outcomes
- Deploy time cut from ~45 minutes to ~4 minutes end-to-end
- 99.99% uptime SLO sustained over the first two quarters
- ~38% infrastructure cost reduction via right-sizing and Spot
- Zero missed regulatory change windows during cutover