Linux skills assessment for hiring: real tasks, real VM, verified per task

The Linux Camp runs hands-on hiring assessments for Linux sysadmin, SRE, DevOps, datacenter technician, and GPU-cloud operations roles. You send a candidate a link. They answer a short block of multiple-choice screening questions on real-world Linux, then get a real Linux virtual machine in the browser that boots in about one second, a ticket-style task list, and a clock. Each task is verified by the machine as they finish it, every command is recorded with a timestamp, and you get a scorecard you can defend: per-question results and per-task verified pass or fail. Real-world Linux skills, not coding puzzles. The pilot is free for your first 25 candidates.

How it works

1. Pick a role pack. Linux sysadmin and datacenter technician are live today as demos; SRE, DevOps, and GPU-cloud operations packs are assembled from the same labs in a pilot.
2. Send the link. The candidate signs in, reads the rules and the time limit, and confirms the honesty statement. Nothing to install.
3. They answer the screening questions (multiple choice, graded on our server, locked on submit), then work on a real machine: a real Linux VM boots in about one second; ticket-style tasks on the left, a terminal on the right, no hints; each task verifies itself as it is done.
4. You read the scorecard: per-question results, per-task verified pass or fail, time used, a shareable result page, and in the pilot the recorded command log and a follow-up script for the interview.

What you get back

The scorecard is the product. Per task: verified pass or fail with the exact check that ran, the points, and the time used. Overall: a percentage and a result page you can share with the rest of the panel. In the pilot, the recorded command log arrives with every scorecard, plus a ten-minute follow-up script built from the candidate's own commands.

Scenario packs by role

• Linux sysadmin (demo live now): ten screening questions on disks, services, permissions, users, cron and networking; then onboard a user, fix a shared folder, lock down a config, bring a broken service back, schedule a job, read a log, free a disk, read the network path.
• Datacenter technician (demo live now): eight screening questions on BMC, SEL, SMART, optics, LLDP and FRU; then BMC address, FRU and serials, failing drive from SMART data, optic serial and switch port, port with no link, BMC event log and power state, a real service fault.
• SRE and senior admin: disk full that du cannot see, runaway process and the OOM killer, DNS and firewall faults, a change with a safety check.
• DevOps and networking: routing and interface state, packet captures that prove a fix, service recovery with persistence.
• GPU-cloud operations: GPU host health from the shell, fabric link state, BMC and out-of-band tasks, escalation write-up.

Hardware facts in a scenario (serial numbers, BMC addresses, event logs, optic data) come from simulated tools on the machine; the Linux system itself is real.

Integrity: evidence, not surveillance

Every candidate gets a unique machine. Every command and its timing is recorded. The hardware facts in a scenario exist nowhere on the internet. A scripted ten-minute live follow-up has the candidate explain their own commands. There is no webcam proctoring, and we do not claim to detect AI use; we make the work reviewable.

Pilot first, then per candidate

• Pilot: free, your first 25 candidate sessions over 60 days, with the Linux sysadmin screen or the datacenter technician check, verified scorecards, result pages, recorded command logs, a setup call and a debrief.
• Per candidate: from $15 per session in packs of 10, 50, or 200; credits do not expire; CSV export of results.
• Team and custom: monthly plans, scenarios built for your stack, GPU-host scenarios, and volume pricing quoted on request; purchase orders and invoices accepted.

Fair by design

Tasks mirror the job, which is the shape of a content-valid work sample. We do not claim EEOC validation or a bias-free test; we give you per-task evidence and exports for your own adverse-impact analysis, extended-time variants on request, a candidate notice and consent screen before the clock starts, and retention you control. We collect no demographics, no webcam, no keystroke biometrics.

Questions employers ask before a pilot

Does the candidate need to install anything or create an account?

Nothing to install: the machine runs in the browser. The candidate signs in with Google or with an email they can verify, so the machine is theirs alone and the result is tied to a real person. In a pilot you send the link; the candidate does not need a subscription or a card.

How are tasks verified, and can I see exactly what the candidate typed?

Each assessment has two parts. The screening questions are multiple choice, graded on our server against a key that never reaches the browser, and locked once submitted. The hands-on tasks run on a real Linux virtual machine with a check script on it: each task is verified from the machine state or the answer the candidate recorded, not from a self-report, and every command is recorded with a timestamp. The scorecard shows per-question results and per-task verified pass or fail with the exact check that ran; pilots receive the recorded command log alongside it.

What stops a candidate from using ChatGPT, Google, or a friend to pass?

We sell evidence, not surveillance. Each candidate gets a unique machine, there are no hints, every command and its timing is recorded, the hardware facts in a scenario exist nowhere on the internet, and we give you a ten-minute live follow-up script where the candidate explains their own commands. There is no webcam proctoring, and we do not claim to detect AI use; we make it reviewable.

Can I set a time limit, and what happens if the candidate runs out of time or loses connection?

Every assessment has a time limit (the demo is 25 minutes). At zero the machine is frozen and whatever is true at that moment is what gets scored. A dropped browser reconnects to the same machine while the clock keeps running. Extended-time variants for accommodations are available on request.

Do we pick from your scenario library, or can you build scenarios for our stack?

Both. Two packs are live today as demos: the Linux sysadmin screen (ten questions, eight tasks) and the datacenter technician check (eight questions, eight tasks). More packs are assembled from labs that already run on the platform: broken-server incidents (disk full, runaway process, service that will not start, DNS, permissions), RHCSA-style administration, networking, BMC and IPMI work, and GPU host health. Questions and scenarios for your own stack are built with you as part of a pilot and priced on request.

How fast do I get results and what does the scorecard include?

Results are instant. The scorecard shows the percentage, the screening questions answered correctly, per-task verified pass or fail with points, time used, and a shareable result page. The pilot adds the recorded command log and a PDF export.

What does a pilot cost, and how does pricing work after that?

The pilot is free: your first 25 candidate sessions over 60 days, with the Linux sysadmin screen, the datacenter technician check, or a pack we assemble for your role. After that, sessions start at $15 per candidate in packs, and team plans, custom scenarios, and GPU-host scenarios are quoted on request.

Do you integrate with our ATS, and how long do you keep candidate data?

Not yet. Result links and CSV come first; Greenhouse and Ashby integrations are next. We store the candidate account email, the commands run, timing, and the results. Retention is set per employer, with a four-year hold option for employers who need to keep automated-decision records under California rules. Candidates see a notice before they start, and deletion is honored on request.

Try it or request a pilot

Run the free demo yourself at the datacenter technician skills check (login required, 25 minutes), or use the inquiry form on this page. A person replies within two business days.