Client names withheld by agreement — the numbers are real.
The situation: A B2B SaaS company was paying for a full rack in a colocation facility with a contract renewal 90 days out. Their 40 VMs and 3 databases had to be out before the renewal date — or they'd be locked in for another year.
What we did: Ran the assessment in week one, then moved in three waves using automated replication. Databases went to RDS with read replicas for the cutover window; traffic flipped over a single weekend with a tested rollback path.
The situation: A payments startup's Java monolith deployed once a month — a full-day, all-hands event. Regulatory pressure meant auditability and rollback speed were becoming compliance issues, not just engineering pain.
What we did: Decomposed the monolith into services on ECS with blue/green deployments, added CI/CD with automated test gates, and built the whole thing in Terraform so environments were reproducible.
The situation: A digital health company ran a patient portal half on-prem, half on a single AWS account with no guardrails. A HIPAA audit was coming, and their architecture wouldn't survive it.
What we did: Built a HIPAA-ready multi-account landing zone first — SSO, centralized logging, encrypted everything, documented controls. Then migrated the portal with a zero-downtime cutover and full audit trails from launch.