skip to content

Interviews on real, broken infrastructure.

Send your candidate a link. They create a free account, and a live Kubernetes cluster boots for them, genuinely broken, with a time box to bring it back. You get a plain-English report on how they did it and the full command log. You do not need to be a DevOps engineer to read either one.

Free plan, no card · Built by engineers who carry the pager

Practicing for an interview? Try the free labs →

candidate@faultybox:~
Recorded interview: the candidate diagnoses a CrashLoopBackOff caused by a bad database hostname, fixes the deployment env var, and the pod returns to Running. Solved in 11:42 of 30 min.
Kubernetes labs
50+
Kubernetes labs
to a live cluster
<1 min
to a live cluster
graded criteria, every report
5
graded criteria, every report
cameras, mics or screen capture
0
cameras, mics or screen capture

How it works on your side

  1. 01

    Pick a lab

    Filter the interview labs by difficulty and time box.

  2. 02

    Send a link

    A live, broken cluster boots in under a minute. Desktop browser only.

  3. 03

    They troubleshoot like it is the job

    AI and Google allowed. Grader scripts check the cluster on submit.

  4. 04

    You read a report, not a number

    Five criteria, each tied to commands in the log. No terminal needed.

Anyone can tell you whether the cluster came back up. Faultybox tells you how.

Every report has a score out of 100, a recommendation, and five criteria:

Fix correctness.
Grader scripts check the cluster after submit, including a re-apply from the manifests.
Diagnosis.
Whether the wrong value was in their own output before they changed anything.
Blast radius.
What the fix cost: guards removed, resources deleted, privileges widened.
Verification.
Whether they re-checked the failing path after their last change.
Root-cause explanation.
Whether their notes explain the chain from cause to symptom.
Card payments time out, and ledger says it is healthy
Demo lab · not in either library
SolvedAdvance with reservations70/100· 22 of 45 min
Fix correctness3/4
Most of the weighted checks pass; at least one fails.
Diagnosis3/4
Followed evidence from symptom to cause. The wrong value appears in their own output before they changed anything.
Blast radius2/4
One major finding. Capped at 2 by the rules; the evaluator may only lower it.
Verification3/4
Checked the symptom, plus one check beyond it that the fix depends on or affects.
Root-cause explanation3/4
Explained the chain from cause to symptom.
Major · 14:02 · live edit on payments-api

They removed the liveness probe from payments-api directly, and the manifests still have it. Worth one targeted question.

Full command log and file diffs in the report

What your candidate actually opens

A symptom, a short brief, and no hint at the cause. Three of the interview labs:

Payments returns 200 OK, with refunds’ response

30 min
Advanced

The payments endpoint answers 200 OK with another service’s response. Nothing is erroring and no alert fired.

The warehouse can’t schedule, and its data must survive

40 min
Advanced

The warehouse pod has been Pending since an infrastructure change, and the records on its volume must survive the fix.

The gateway can’t reach identity, and nothing was deployed

45 min
Advanced

The api-gateway started failing overnight with no deploy, no config change and no restart.

What changes in your loop

This replaces one stage, usually the take-home or the first technical screen. Everything else about how you hire stays yours.

Today

  1. 1Screen a resume that says "Kubernetes" and hope
  2. 2Book 45 minutes of an engineer to ask trivia
  3. 3Send a take-home and wait a week for it
  4. 4Argue in a debrief about whether the answer was good
  5. 5Find out for real during the first incident, six weeks in

With Faultybox

  1. 1Send a link, no scheduling call, no engineer time
  2. 2Candidate fixes a real fault in a browser, time-boxed
  3. 3Report arrives in plain English with the full command log
  4. 4Debrief on evidence, with the command log open
  5. 5Find out before the offer, not after

Next to HackerRank, Codility, CoderPad and TestGorilla

Good tools for what they test. Faultybox covers a different question: can this person fix a broken cluster, with the command log to show how they did it.

Is this for you?

A fit if

  • You hire for roles that carry a pager: SRE, platform, DevOps, or backend engineers going on-call.
  • Your engineers use AI on the job, and you would rather measure that than police it.
  • You want to see how someone worked, then make the call yourself.
  • Your stack runs on Kubernetes.

Not a fit if

  • You need proctoring, webcams or AI detection.
  • You want one score or a percentile to sort candidates by.
  • You are hiring for a stack that is not Kubernetes yet. Linux, Docker, Terraform and CI/CD are on the roadmap.
  • You want the tool to make the hire or no-hire decision.

What a candidate is actually tested on: the DevOps skills assessment.

Still unsure? Email support@faultybox.com.

Start free →

The questions buyers actually ask

Nobody on my side is a DevOps engineer. Can we still run this?
That is the case we designed for. Every lab states in plain English what it puts under pressure, and you filter by difficulty and time box rather than by tool. The report comes back with the technical judgment already made, so a recruiter or a hiring manager can read it without a terminal.
What stops them cheating?
Nothing, and that is deliberate. AI is allowed, so there is nothing to hide and nothing to detect. Solved or not solved comes from scripts that check the cluster itself, including whether the fix survives a re-apply from the manifests. A memorized walkthrough does not survive that, and the command log shows whether they ever looked at the thing they changed. Randomized fault variants per invite are on the roadmap.
How much does it cost to try?
Nothing. The free plan is 3 interviews and 3 trial runs on every interview lab, with no card. After that you buy interviews in packs from $350, and an interview is only used when the candidate starts. Paid packs are not on sale yet.
All questions →

Run one yourself first. It takes about 30 minutes.

Work a broken lab yourself, read the report it produces, then decide whether you want a candidate in there.

Free plan: 3 interviews and 3 trial runs on every interview lab, no card. Paid packs are not on sale yet.