About
Nonagon Media is an infrastructure and platform engineering practice. We build and run the unglamorous machinery that lets software ship reliably: pipelines, cloud architecture, containers and orchestration, monitoring that actually notices, and the documentation that makes all of it survivable.
We work with teams that need senior infrastructure engineering but do not need — or cannot yet justify — a full-time platform hire. Some are just getting set up. Some have a platform team that is underwater. Some have inherited an estate that nobody left in the building fully understands.
How We Work
A named engineer, not a ticket queue. You deal directly with the person doing the work. There is no account manager in between, and no handoff to somebody junior with a runbook.
We look before we touch. Audits run read-only. Nothing in your estate changes while we are still working out what is there.
Documentation happens as we go, not at the end. It is a deliverable, not a courtesy. If the engagement ended tomorrow your team should be able to operate what we built from what we handed over — that is the test we hold ourselves to.
We tell you when there is nothing to find. Sometimes an estate is already tight and an audit turns up very little. We will say so. That is a worse sales call and a better answer, and it is the only way the good news means anything.
Remediation is quoted separately, from the findings. There is no obligation to buy the fix from the people who found the problem. If you would rather take the plan and build it yourself, that is a perfectly good outcome.
We build the smallest thing that solves the problem. Complexity is a liability you pay for every month, in a bill and in the time it takes the next engineer to understand what they inherited.
Our Principles
Nine sides. No shortcuts.
Automation by default. If a task must be done twice, it belongs in a playbook or a module. If it must be done three times, it belongs in a pipeline. Manual intervention is not a workaround — it is a bug.
Linux at the core. Containers are Linux. Kubernetes is Linux. The cloud is someone else's Linux. When the abstraction leaks, and it always leaks, you need to know what is underneath.
Right-sized engineering. Complexity is not a badge of honour, it is a liability. The right tool is the simplest one that solves the problem. Over-engineering is technical debt with better marketing, and an over-provisioned estate is the same mistake arriving as a monthly invoice.
Finish the work. Be thorough and efficient, and always finish. Finished does not mean working — it worked for whoever built it. Finished means somebody else can operate it from what was handed over.
Document for the next engineer. The next engineer is usually your engineer, and sometimes it is you. Handing over a reorganised system with nothing to explain it is how it drifts back within a year.
Verify, do not assume. Finding an idle-looking instance takes seconds. Establishing that it is genuinely idle and not a quarterly batch job is what takes the day. A green build means the image was pushed, not that the page renders.
Notice before the client does. A service can look alive and be dead. A unit stays active while the thing it runs has stopped, and nothing pages anybody because the process is technically fine. Monitoring has to watch the output, not the process.
Least privilege, every time. Credentials live in a secret manager, not in an environment variable on a command line where the process list can read them. Access is scoped to the job and revoked when the job is done.
Radical transparency. We publish our failures alongside our work — the permission nightmares, the DNS bug that only appeared under load, the silent outage that monitoring should have caught and did not. The best engineering happens when people stop pretending things work the first time.