Cloud migration for a fintech platform

Context

A fintech company processing thousands of daily transactions ran their entire platform on bare-metal servers in a single data centre. Deployments were manual, took four hours of coordinated effort, and required a maintenance window every time.

The team had grown from 3 to 25 engineers. The infrastructure hadn’t kept pace.

Challenge

Migrate 14 interconnected services to Kubernetes without disrupting active transaction processing. The system couldn’t afford downtime — real money moved through it every minute.

The existing services had no containerisation, no infrastructure-as-code, and inconsistent configuration across environments.

Approach

We worked in three phases:

Phase 1 — Containerise and standardise. Each service got a Dockerfile, health checks, and environment-based configuration. We built a staging cluster that mirrored production topology.

Phase 2 — Parallel run. Services ran simultaneously on bare metal and Kubernetes. Traffic gradually shifted using weighted routing. Each service was validated independently before the next moved.

Phase 3 — Cut over and decommission. Once all services ran stable on Kubernetes for two weeks, we decommissioned the bare-metal servers and handed over runbooks to the team.

Outcome

Deployment time dropped from 4 hours to 12 minutes
Zero downtime during the entire migration
Infrastructure costs reduced by 35% through right-sizing
Team ships independently — no more coordinated deployment windows
Full infrastructure defined in Terraform, versioned and repeatable

Technologies