Financial Services Firm

Every server, database, and network rule had been created by hand and existed only in one engineer's memory. The team knew a DR event would be catastrophic — they just hadn't had time to fix it.

We started with a discovery sprint: mapping every running resource, documenting its purpose, and classifying it by criticality. That gave us the order of work. We didn't try to IaC the whole estate at once; we started with the stateless compute layer, validated the Terraform modules against the existing state, and imported live resources before touching anything.

After six weeks the entire infrastructure was described in code and reproducible from scratch. The seventh and eighth weeks were DR design: a secondary region, automated snapshot replication, and a tested failover runbook. We ran the first DR test with the client's team — simulated primary region failure, executed the runbook, and measured time to recovery.

First attempt: passed. Recovery time met the internal SLA the team had set for themselves.