Agentic DevOps

The short story is that we’re building a software factory where autonomous AI agents actually code and ship our features. Since the code is driven by AI, our infrastructure needs to change with it.

This is a hybrid DevOps role. Half of the job is what a senior DevOps engineer has always done: keeping cloud infrastructure healthy, deployments smooth, costs sane, and the on-call rotation quiet. The other half is newer and more interesting: supporting a production multi-agent AI pipeline alongside our AgentOps engineer, owning the infrastructure layer beneath autonomous agents that ship our code.

You will not be the person tuning prompts or rewriting agent SOUL files - that lives with AgentOps. You will be the person who makes sure the EC2 instances, EKS clusters, Lambda functions, and ArgoCD pipelines that the agents depend on are rock-solid, observable, and ready when they need them.

What we're really looking for?

We're looking for an infrastructure expert who views Terraform and IaC as a foundation, not the end goal. While keeping our infrastructure-as-code rock solid, you'll spend your time diving deep into system internals, solving complex production bottlenecks, and architectural challenges where DevOps meets AI.

You think infrastructure-first.

When a production system has a problem, your first questions are about networking, IAM, capacity, deployment state, secrets. You can read a CloudWatch dashboard the way most developers read a stack trace.

You're comfortable with both worlds.

You can spend the morning debugging an EKS pod that won't schedule and the afternoon helping the AgentOps engineer figure out why a cron job on EC2 is missing its window. Neither feels foreign.

You believe in GitOps.

Infrastructure is code. Changes go through pull requests. Rollbacks should be a git revert . Manual clicks in a console are an emergency, not a workflow.

You're paranoid about blast radius.

Before you change a security group, you know exactly what depends on it. Before you redeploy, you know which services will recycle. You think in terms of "what could break, and how do we recover?"

You write narrow, deterministic automation.

You prefer one well-tested Bash script over a sprawling Python framework. You don't build abstractions for hypothetical future needs. Your YAML is boring on purpose.

You're an on-call professional.

You've owned production. You know what a good runbook looks like, what a useless alert looks like, and how to write the difference. Quiet pagers are an art form.

You're curious about AI infrastructure as a category.

You don't need to be an ML engineer, but you find the systems-level challenges of running autonomous agents at scale interesting - the unusual cost shapes, the long-running stateful sessions, the deterministic-vs-probabilistic monitoring problem. You'd want to learn this layer.

Experience we expect

• 5+ years in DevOps / SRE / platform engineering with hands-on production ownership.

• Deep AWS experience - EC2, EKS, Lambda, IAM, VPC networking, CloudWatch, Secrets Manager. You should have opinions on when to use each, and where AWS's defaults will hurt you.

• Production Kubernetes, ideally EKS. You can write a Deployment, debug a CrashLoopBackOff, configure HPAs, and reason about resource limits. You understand what kubectl is actually doing under the hood.

• GitOps with ArgoCD (or Flux). You've built ArgoCD applications, dealt with sync failures, and know when self-healing is helping vs hiding problems.

• Infrastructure as Code - Terraform, CDK, or Pulumi. You've maintained a non-trivial codebase, not just bootstrapped one.

• CI/CD pipelines at production scale - GitHub Actions or similar.

• Strong Linux and shell skills - cron, systemd, log parsing, subprocess management. Python or Go for tooling beyond shell.

• Observability stack experience - Prometheus + Grafana, Datadog, or equivalent. You can build dashboards that engineers actually use, and alerts that don't lie.

Bonus

• Experience supporting AI / ML production systems (model serving, GPU clusters, batch inference pipelines, vector DBs).

• Familiarity with multi-agent orchestration platforms or LLM API ops (rate limits, token cost monitoring, prompt caching).

• Experience with Discord / Slack ops integrations - bot infrastructure, webhook reliability, channel monitoring.

• Background in cost optimization at the AWS account level.

• Fluent in English.

What makes this role different

• You work with the AgentOps engineer, not under them. You own the infrastructure layer; they own the agent layer. You meet in the middle, and the boundary is collaborative, not territorial.

• You're closer to the new stack than most DevOps roles. While other DevOps engineers are still being asked to support traditional microservices, you'll be supporting a production AI pipeline - a category that's still being defined.

• Your work is visible. When deployments go smoothly, the agents ship more tickets. When the EKS cluster is healthy, the engineering team forgets about it. Both outcomes are measurable.

• You'll shape the playbook. "How do you operate infrastructure for an autonomous agent team" is not a solved problem. You'll write the answer.

Who this role is not for

• People who want to write only Terraform all day. The job is broader.

• People who want to stay away from AI workloads. You'll be supporting one of the most active AI-driven workloads in the company.

• People who need a perfectly defined scope. The boundary between DevOps and AgentOps will keep shifting as we learn.

• People who prefer enterprise-process environments over pragmatic ownership. We move fast and own outcomes.