Roku · 2020–2021
Attributing $10–12M a quarter of AWS spend, mid-migration
A cost-attribution pipeline for AI/ads forecasting workloads, built while co-leading the move from ECS/EC2 to Kubernetes. Cost data treated as an engineering artifact, not an accounting afterthought.
Headline
$10–12M
AWS spend attributed per quarter
ECS → K8s
migration co-led
Istio · ArgoCD
blue/green delivery
Multi-cloud
Azure expansion
The problem
AI and ads forecasting workloads were growing fast, and the quarterly AWS bill was large enough that “infrastructure” as a single line item hid real decisions. Teams couldn’t see what their own workloads cost, so nobody could be asked to change them.
What I built
A pipeline that tied spend back to the workloads and teams generating it, so the bill could be read as a set of owned costs rather than one number. It ran while the platform itself was moving from ECS/EC2 to Kubernetes with Istio, Grafana, blue/green deploys through ArgoCD, and an expansion into Azure, which meant attribution had to survive the ground shifting underneath it.
The lesson I kept
Attribution is only useful when it lands with someone who can act. The migration and the cost work turned out to be the same project: every new deployment path was also a new place for cost to disappear if it wasn’t tagged from day one.