San Francisco, CA · Open to DevOps & platform roles
Senior DevOps Engineer
Six years building AWS infrastructure, Kubernetes platforms and CI/CD automation. I build the shared delivery tooling that 18 production services across 7 application teams ship through — EKS, Terraform, Helm, GitHub Actions and Argo CD.
And I build with AI rather than just using it: Claude Code Agent Skills and MCP integrations for infrastructure, deployment validation and troubleshooting.
Delivery
Shared delivery tooling for 18 production services across 7 application teams. Reworked CI/CD to build the artifact once and promote that same image through every environment, with consistent approval and rollback behaviour across all of them.
~40% fasterdeployment lead time
Infrastructure
Replaced manual provisioning with declarative resources the application team owns in its own repo. Standard service setup dropped from about three days across three queues to a single pull request.
~3 days → <4 hoursstandard service setup
Verification
Added deployment safeguards that catch AWS authentication failures, incomplete rollouts and incorrect releases before they reach production — rather than after someone notices in the dashboard.
The principle underneath all of it: the only trustworthy signal is the one that ties the running workload back to the artifact you actually built. A pipeline that reports success without checking that is measuring the service, not the release.
Reliability
Built the Prometheus/Grafana observability baseline with SLO-based alerts and Alertmanager routing. Earlier, at AppsFlyer, I tuned alerting for Kubernetes and Kafka workloads down from ~30–40 pages a week to about 15 — without losing detection of real incidents, which is the only version of that number worth quoting.
~35% fewernon-actionable production alerts
Scale
Reusable Terraform for AWS and EKS environments used by three application teams took new-environment provisioning from ~2 days to under 2 hours. Rightsizing and autoscaling across six processing services cut EKS worker-node spend while holding latency and throughput targets.
~$18K/mo (~15%)EKS worker-node spend removed
Intelligence
Claude Code Agent Skills give repeatable Terraform, deployment-validation and troubleshooting workflows a fixed shape instead of an open prompt, and MCP integrations connect the nine-person platform team to GitHub, Kubernetes, AWS and operational tooling.
Reads run locally per engineer; anything that changes infrastructure goes through one path, with a person approving the change. The boundary is RBAC and quotas, not wording.
Stack
Experience
Contact
I’m looking for a DevOps or platform team that takes fundamentals seriously and is genuinely putting AI into how engineers work — with the guardrails to do it safely. Based in San Francisco, open to relocation.