The problem
Lift-and-shift was costing them ₹14L/month in EC2 alone. We re-architected to a microservices pattern on EKS with reserved instances, savings plans and right-sized workloads, then layered observability on top so the team could see costs in real time.
Our approach
We started with a 1-week read-only audit of the production environment, billing data and deployment pipeline. The audit produced a single-page proposal: quick wins (this sprint), structural fixes (this month), and strategic moves (this quarter). The client picked the sequence; we executed in two-week sprints with weekly demos.
Stack we used
AWS EKSTerraformPrometheusGrafanaGitHub Actions
What changed
- Infrastructure rationalised against actual workload patterns — no over-provisioning.
- CI/CD pipeline standardised so every service deploys the same way, reducing "works on my machine" cases.
- Observability added before any change rolled out — we measured the delta, not guessed at it.
- Documentation and runbooks handed over so the in-house team owns the system end-to-end.
What the founder said
"We expected savings. We didn't expect to also get a clearer view of where our infrastructure was actually spent. The team became part of our planning conversations, not just an outsourced ops vendor."
Founder, anonymised — full name available under NDA.