Forget quizzes and abstract tests. Every candidate fixes a real outage in a real environment, and your experts only meet the finalists.

30-min demo

AI changed what a test proves.

  • The quiz

    Solved by AIA model answers it in seconds. You end up grading the model, not the engineer.

  • The take-home

    UnverifiableNobody watched it being done. Hard to tell who, or what, did the work.

  • The CV

    Is no proofIt says Kubernetes. It does not say whether they can bring a cluster back.

  • The interview with your experts

    Works, at a costYour experts give an hour to every applicant, including those they rule out at once.

  • The work itself

    What still proves somethingOn a real infrastructure, recorded from the first command to the last.

You send a link. You get evidence.

Three steps to the shortlist, without taking up a single one of your experts.

The real job, on real infrastructure.

A standard interview reproduces neither the job nor the infrastructure. Here, each candidate works in a real environment, from their browser: you see how they work before you hire them, not after.

A decision you can defend.

What they did, how they went about it, where they stand. No debrief meeting, no engineer at your side.

Passed or not, at a glance

85 / 100 Above the threshold

Threshold 70Stage median · 75

3rd of 9 candidates at this stage

What reassures

The queue is built on the team's module

The table follows the key the worker expects

What worries

No alarm on the dead-letter queue

Everyone compared, on the same mission

Directs and checksTargeted questionsDelegates and checksAI not usedDelegates, no visible check2030405060708090100median · 75pass score · 70Validator score →Sarah85 · Directs and checksHugo75 · Directs and checksLéo95 · Targeted questionsInès60 · Targeted questionsEva80 · Delegates and checksTom70 · AI not usedNora45 · AI not usedKarim90 · Delegates, no visible checkMalik30 · Delegates, no visible check

The session, replayed

1×
Journal

Extract of a real session · preproduction

How they work with AI

Directs and checks

3 requests · 2 precise, 1 question

PreciseAdd the team's tags to both queues.

QuestionWhich key does the worker use for the orders table?

PreciseAdd an alarm on the dead-letter queue.

What they looked up

History · 14 pages on 4 sites · 3 searches

Rregistry.terraform.io6 pages · 9 min

Ddocs.aws.amazon.com4 pages · 5 min

Eeu-west-1.console.aws.amazon.com3 pages · 4 min

⌕“terraform sqs redrive_policy”Search · Google

You decide. The email is already drafted.

Sarah N. 85 / 100 · Directs and checks

RejectPut on holdAdvance to “Live interview” →
Email to the candidate · drafted by AI, editableFR · EN[Acme Conseil] Next: a live interview

Hello Sarah, Thank you for your work on the mission. Your order queue is built on the team's module, which is exactly what we hoped to see. One point to go through together: the alarm on the dead-letter queue. We would like to continue with a live interview. Camille · Acme Conseil

Advance and send →

Example · fictitious data

AI is allowed. Blind trust goes on the record.

Your engineers work with AI every day. So do your candidates. UnRank does not ban it: it shows you how each of them uses it.

  • You choose, stage by stage: a built-in AI agent, or no AI.
  • Every request to the agent is logged and classified in the report.
  • Using the agent costs no points: the score comes from the checks alone.
  • AI analyses decide nothing. You do.

Missions that look like the job.

A generic test filters on trivia nobody uses at work. A mission close to the job filters on what your hire will do from day one. Twelve are ready: no test to write.

The status page went dark the night before a sale. Nobody can reach it any more.

EC2IAMS3

9 fixesto repair5 guardsnot to break

The config bucket stays private

Junior · 60 min01 of 12

Or your own: your production becomes their mission.

You describe the role and your architecture. We place the fault, what has to be built and the automatic checks on it. Only you see it in your catalogue.

Get my custom mission

Private missionThe invoices stopped comingSenior · 75 min

Public API4Orders serviceCacheEvent queueBilling worker1Customers database3Dead-letter queue2
  1. 1The faultThe billing worker has been falling behind since Tuesday's release.
  2. 2To buildA dead-letter queue for failed messages.
  3. 3GuardNo invoice issued twice.
  4. 4GuardThe public API stays up.

Example · fictitious data

What changes for you.

IT services and consulting

What it costs you todayYour client runs the technical interview again.Your vetting proves nothing to them. A rejected profile is a consultant left on the bench.

What you gain with UnRankYour client watches the consultant work.They approve on evidence, not on a CV, before they even meet.See the IT services page

Recruitment agencies

What it costs you todayThe client's technical test decides for you.A candidate who fails at the client is a lost placement, and your fee with it.

What you gain with UnRankYour candidates have already proved themselves.You present the work done on a real mission, not just a CV.See the recruitment agencies page

Scale-ups and software companies

What it costs you todayYour seniors pay for every hire in interview hours.Including for candidates they would have turned down at once.

What you gain with UnRankYour seniors only see the finalists.Screening runs without them, on a real mission. Their hours go back to the product.

Hire on evidence

Your questions, straight answers.

Another question? We reply within one business day.

Ask a question
01Which roles and technologies do you assess?

DevOps, SRE, cloud and platform engineers. Twelve missions at launch, 60 to 90 minutes each: AWS, Kubernetes, Terraform, Docker, Linux, Ansible, Azure, PostgreSQL, Kafka.

02We already have a technical test. What does this change?

A quiz or an algorithm exercise scores an answer. A take-home comes back without showing how it was done. Here the candidate works on real infrastructure, the result is checked on that infrastructure, and you replay every step they took, requests to the AI included.

03Can the candidate use AI?

You decide, stage by stage: a built-in AI agent, or no AI. With the agent, every request the candidate makes is in the report, classified. The score comes from the checks alone: using the agent costs no points.

04How do I spot cheating?

Screen sharing is mandatory and the whole session is recorded. The report flags what deserves a look: a paste close to the solution, a search for the answer key, a screen share that stopped. Searching the web is normal and is not flagged.

05What does the candidate go through?

They get a link, check their camera and microphone, accept the recording, then work in their browser: nothing to install. The timer starts only once they are in the mission. They receive your decision by email.

06Where does candidate data go?

You remain responsible for it; UnRank processes it on your behalf, under a data processing agreement (DPA) included in the terms. The platform is hosted in the European Union, no data is sold, and the AI analyses take no decision.

Read the privacy policy

07How do I get a custom mission?

We build it with you, from your production and the role. It appears in your catalogue only. We scope it during the demo.

08How much does it cost, and how do I start?

During the private beta, the offer fits your hiring volume: there is no public price. In 30 minutes we show you a real mission and a real report, then open your workspace. Prefer to write? You get a proposal within one business day.

Your engineers have better things to do than your interviews.

Stop having them write and maintain tests. We reproduce your production, as closely as possible.

We answer within one business day.

Enter a valid email, for example name@company.com.

Book a demo