/cloud-finops
Cloud cost is an engineering problem
The biggest savings come from architecture, not discounts. I make cost visible to the engineers who own it, then change the systems that drive it.
Architectural invariants
- Cost is tracked per service, next to latency and errors.
- Savings come from architecture, never from reliability.
- Cross-zone traffic is designed, not accidental.
- A team that can't see its spend can't own it.
Proof in production
- Built a near real-time serverless cost pipeline (AWS Lambda, Athena, Python) that showed every service owner their spend, saving over $1M a year.
- Redesigned cache serialization for another $1M+ a year.
- Cut inter-AZ traffic costs and enabled Spot scaling through zone-aware routing.
- Set up real-time per-service cost tracking as a platform capability.