Cut cloud spend 38% in 6 weeks.
Multi-region platform was burning compute. We re-sized the fleet, moved to Graviton, and added a savings plan model that paid back in 21 days.
What we walked into.
Bills were climbing 8–12% month-over-month while revenue growth had flattened. EC2 represented 64% of spend across three regions. The team suspected over-provisioning but had no time to investigate properly.
Previous attempts at rightsizing had been done in isolation and rolled back after a performance incident. Making the team understandably cautious about changing instance types.
How we shipped it.
Six weeks of metric collection first
Before touching anything, we instrumented every workload with proper CloudWatch agent metrics, captured a 30-day usage profile per service, and identified the actual CPU/memory ceiling per workload rather than guessing.
Graviton migration for stateless workloads
Migrated 12 stateless services to Graviton-based instances. 20% cheaper at the instance level, with no observable latency difference at the application tier.
Savings Plan model matched to usage shape
Modelled three Savings Plan scenarios against the cleaned-up baseline. Recommended a 3-year compute SP at the 60th-percentile baseline rather than the typical 80th. Accepting some on-demand burn in exchange for flexibility.
Dev/staging onto Spot
Consolidated dev and staging into a single account on Spot Fleet with interruption handling. 70% cheaper than the original dedicated environments.
AWS services in this engagement
EC2 (Graviton)
Lambda
RDS
S3
CloudFront
ALBWhat shipped.
Monthly AWS bill down from $33,600 to $20,800. Annualised, that is $154k saved on a six-week engagement.
The Savings Plan model has held up across three quarters of growth. The team now runs a quarterly cost review using the playbook we wrote.
Hardened 47 IAM roles in 11 days.
Series B fintech needed audit-ready IAM before SOC 2. We refactored every role into least-privilege Terraform modules.
ArchitectureMigrated from Heroku to a cloud platform in 9 weeks.
B2B platform hit Heroku ceilings. We landed them on ECS with managed Postgres, zero downtime, and a runbook the team actually uses.
SecurityClosed 12 audit findings before pen-test.
Health-tech client needed to ship the audit. We instrumented CloudTrail, tightened S3, and resolved every critical inside two sprints.
Tell us what you're trying to ship.
A 30-minute scoping call with the engineers who would do the work.