Linux System Administrator

Before you apply

• We're direct about how we work. Read this section first.
• This is a small team and you will own real systems. There is no tier one to escalate to. When something breaks, you are the person who fixes it, and you will have the access and the authority to do that.
• We value people who know what they don't know, and say so. If you're honest about your gaps, we'll invest in closing them. If you hide weaknesses or oversell your experience, you won't last here.
• Most of this job is preventing work, not doing it. The measure of a good month is that very little happened. If you want a role where you are visibly firefighting, this is not it.
• We move fast. Growth happens outside comfort zones. If this sounds harsh, we're probably not a fit. If it sounds like the work you have been waiting for, keep reading.

The role

Allerin builds AI and IoT products. Edge to cloud video intelligence, computer vision systems, and data platforms. We also build and run software for clients who measure whether it works. All of it runs on infrastructure somebody has to keep standing up, and that somebody is this role.

The estate is self-hosted rather than rented. Proxmox virtualisation on our own hardware, a fleet of Ubuntu VMs, a self-hosted GitLab with its own CI runners, PostgreSQL, reverse proxies and certificates, application services under systemd, and scheduled jobs that have to actually run. Some of it faces the public internet. Some of it is client infrastructure where an outage is their outage.

What you'll own

• The fleet: provisioning, patching, and knowing what is running where without having to go and look.
• Keeping it up: the services, the proxies, the certificates, the timers. Noticing a disk filling before it fills.
• Backups you have actually tested. A backup nobody has restored from is a belief, not a backup. Proving that is part of the job and we will ask you about it.
• The network end: DNS that resolves, firewall rules somebody can read, and understanding what happens when a name does not resolve at three in the morning.
• Automating yourself out of the repeat work: scripts, configuration management, whatever fits. We would rather you spent a day removing a task than a year doing it.
• Writing down what you did. The next person, who may be you in eight months, needs to be able to follow it.

What you must arrive with

We do not require a certification and we do not count years.

• You run Linux properly. Not just commands you have memorised, but knowing where to look when the obvious answer is wrong.
• You can debug something you did not build. Logs, processes, disk, network, in a sensible order.
• You script. Bash at minimum, Python or similar preferably. Enough to stop doing a thing by hand.
• You understand what happens between a browser and a service, well enough to find which part is broken.
• You have restored something from a backup, for real, under pressure or in a drill.

A GitHub account is required to apply. We are not counting stars. Scripts, dotfiles, configuration, a tool you wrote to stop doing something twice, anything that shows how you work. A small, honest repository is fine.

Nice to have

• Proxmox, KVM, or any virtualisation you have run rather than clicked through
• Self-hosted GitLab or similar, including the parts that break
• PostgreSQL administration, and knowing when a problem is the database rather than the disk
• Configuration management, Ansible or otherwise
• Monitoring and alerting you built rather than inherited
• Certifications, and we will read them, but they will not decide anything on their own

What we offer

• Real ownership of real systems, not a ticket queue
• Hardware and an estate you can actually change
• A team that would rather you fixed the cause than the symptom
• Investment in your growth if you're honest about where you need it

How we hire

Apply through the form. Seven questions, six of them required, and they are about things you have done.

If we talk, we will describe something that is broken and watch how you narrow it down. There is no single right answer and you will not be asked to recite commands.

A human reads every application, and answers that read as model-generated are tested in depth in that conversation, where they do not survive.