Responsibilities
Both tenancy models, as one system. Terraform, Helm, and ArgoCD delivering to shared and dedicated environments alike. You own the pipeline and the guardrails on it: progressive rollout, blast-radius containment, and drift detection that catches a mistake before a customer does.
Efficiency and cloud cost. Dedicated infrastructure per customer makes unit economics an engineering concern, not a quarterly finance exercise. You own attribution, right-sizing, spot strategy, and autoscaling that actually scales down - and you defend efficiency as a design constraint on every new deployment.
Turning operations into product. Onboarding, version rollouts, and config changes are tickets today; they should be self-service and boring. You decide what gets automated, what goes back to the owning team, and what still needs an expert.
Scale under real load. Temporal, KEDA queue-based autoscaling, and Karpenter are live and still maturing. You own how the platform grows and shrinks against real traffic, and what that costs per customer.
Security engineering for regulated customers. KMS-based key management, SIEM pipelines, WAF policy, certificate and secret lifecycle, per-customer SSO, vulnerability remediation, and turning penetration test findings into shipped fixes. Our customers are banks, insurers, and health funds. Their auditors are effectively part of our roadmap.
AI-driven operations. Our infrastructure repositories are already built for agentic work, with an internal harness of shared skills and agents. You inherit it, extend it, and push it into places it does not reach yet: triage, rollout verification, cost anomaly investigation, log analysis, and the long tail of operational work that has never been worth a script.
Reliability of what you ship. You are on call for your own systems, paged through PagerDuty against alerts you own.
Requirements
6+ years in DevOps, including production ownership of Kubernetes at meaningful scale. EKS preferred.
Operating a fleet, with infrastructure as code. You have run many similar-but-not-identical environments with Terraform, and felt what happens when a change has to reach all of them. You have designed modules other people build on, and you have recovered state when it went wrong.
Deep AWS and Cloudflare as a production edge. You have run AWS at the account, network, and IAM level rather than consumed it, and you have operated Cloudflare in production: WAF policy, origin protection, DNS, and rate limiting.
You treat cloud cost as an engineering metric you own. You have measured it, attributed it, and brought it down - through right-sizing, spot, autoscaling, storage lifecycle, or architecture - and you can talk about what you saved and what it cost you in complexity or risk to save it.
Python and Bash as an engineer, not a scripter. Much of the leverage in this role comes from tooling you write yourself.
You have served regulated enterprise customers. Financial services, insurance, or healthcare - environments where penetration tests, security questionnaires, encryption requirements, and customer-imposed controls shape what you are allowed to build. You know how to satisfy an auditor without freezing the platform.
Genuine AI fluency in how you work. You use agentic coding tools (Claude Code, Cursor, or equivalent) as a core part of your engineering practice, and you have automated real operational work with them. You have a considered view on where to trust an agent in production and where not to, and you can defend it.
CI/CD across a mixed estate. Jenkins, CircleCI, and GitHub Actions all run here. You have owned pipelines in more than one of them and have an opinion about what belongs where.
Open-source observability you have operated, not just read. Prometheus, Grafana, and Loki as systems you have run, scaled, and tuned.
Proactive ownership. You find the problem before it is assigned to you, and work you pick up does not come back. On a team this size, nobody is tracking your queue for you.
Strongly preferred - any of these moves you up the stack:
Temporal.io in production: worker scaling, queue backpressure, and operating it as infrastructure rather than consuming it as a library.
Event-driven autoscaling with KEDA, and node lifecycle management with Karpenter, tuned against real cost and latency targets.
Secret management at fleet scale: SOPS, KMS, key rotation, and hold-your-own-key models where the customer holds the key material.
Multi-region production operations, including data residency and per-region compliance obligations.
Building internal developer platforms or self-service tooling that measurably reduced how often engineers had to ask for help.
Migrating source-control or CI/CD estates, or consolidating fragmented build tooling onto a single modern platform.
Kubernetes network and workload security in depth: CNI policy with Cilium or Calico, service mesh, and enforcing isolation between tenants at the network layer.
Startup instincts plus enterprise discipline. You have worked somewhere small enough that you owned the whole problem, and somewhere serious enough that "move fast" had a blast radius.
Who you are
