Hands-on technical assessments

Hire the engineer who can actually fix it.

Candidates get a real, deliberately broken environment and a clear task. They investigate, repair and validate it the way they would in production, while you get the evidence.

Request a demo Browse the lab catalogue

Every Kubernetes lab includes

A fresh, isolated cluster created for the candidate and destroyed afterwards
A browser terminal with kubectl and the tools the task needs
Node-level shell access for kubelet, runtime and control-plane labs
A time limit and automated validation on submit
Recorded command history for reviewers

For hiring teams

Better signal, with less of your engineers' time

Practical assessments filter early and give interviewers something concrete to discuss.

See real working method

How the candidate gathers evidence, forms hypotheses and narrows the fault.

Compare consistently

Same environment, same fault, same checks for every candidate on a role.

Save senior time

Your engineers interview fewer, better-qualified candidates.

Decide on evidence

Check results and command history, not a general impression.

Workflow

How an assessment runs

01
Define the role
Pick labs from the catalogue for the role and level, or ask us to recommend a set.
02
Invite the candidate
The candidate receives a private link. No account or software is needed.
03
Candidate works the problem
A fresh environment is created with the fault injected. The timer starts when they begin.
04
Review the results
Automated checks run on submit. You receive scores, evidence and reviewer notes.

Roles and difficulty

Matched to the seniority you are hiring for

Labs are levelled internally. Candidates see the task, not the difficulty rating, so the level does not change how they approach it.

Junior
Reads errors and events, fixes common workload faults, and follows a methodical process.
Mid-level
Diagnoses across components (DNS, Services, storage, RBAC) and validates the fix independently.
Senior
Works at node and control-plane level, handles ambiguity, and avoids collateral damage.
Lead / Architect
Designs for failure: availability, isolation, security controls and safe operational change.

Roles covered: DevOps Engineer, Platform Engineer, SRE, Kubernetes Engineer, Cloud Engineer, CI/CD and Release Engineer, AI Infrastructure Engineer.

Lab catalogue

Example lab scenarios

Ready-made labs run from our library today. Tailored labs are example scenarios we build around your stack. Every area is also covered in expert interviews.

Ready-made labTailored lab

Ready-made only

Showing 53 of 53 labs

How automated validation works

When the candidate submits, hidden checks inspect the real state of the environment. They do not look at what the candidate typed; they test whether the system works.

Each check tests one outcome, such as “the Service name resolves from the affected Pod”
Checks carry points; a pass mark is set per lab
Guard checks catch shortcuts such as deleting the broken workload
Every lab is tested against a clean, a broken and a repaired environment before release

How human review works

Automated checks show whether it works. A reviewer looks at how the candidate got there, because two passing results can reflect very different engineers.

Review of command history: investigation order, risky actions, validation steps
Rubric scores for method, accuracy, safety and validation
Written comments your hiring manager can read in two minutes
Optional follow-up questions for the live interview

Candidate experience

Fair, clear and respectful of their time

Strong candidates often have several offers. The assessment should feel like real work, not an obstacle course.

Browser only

No installs, VPNs or local clusters.

Clear task

A written brief that states what success looks like.

Visible time limit

Most labs take 15 to 40 minutes.

Private and isolated

A dedicated environment that is destroyed afterwards.

Combined assessment + interview

One report that brings lab evidence and interview together

For senior and lead roles, the strongest signal comes from a lab followed by a focused conversation about what the candidate did and why.

1Candidate completes a hands-on lab.
2Your team reviews the results and activity.
3An expert interviewer probes decisions, trade-offs and alternatives.
4You receive a consolidated evaluation and recommendation.

Discuss your hiring requirements

Evaluation report

Candidate A · Platform Engineer (Senior)

Sample report

Hands-on lab

86 / 100

Live interview

4.0 / 5

Recommendation

Proceed

Troubleshooting method

Reproduced the symptom, then narrowed scope methodically.

Technical accuracy

Correct root cause; fix limited to the faulty configuration.

Validation

Proved the fix from the affected workload before submitting.

Communication

Explained decisions clearly; missed one follow-up risk.

Attached evidence: automated check results, terminal command history, interviewer notes.

FAQ

Common questions

What does the candidate need?

A modern browser and a stable connection. The terminal runs in the browser; there is nothing to install and no credentials to share.

Which areas have hands-on labs?

All of them. Kubernetes and Kubernetes security have 33 ready-made labs that run from our library, covering workloads, networking, nodes, the control plane and etcd. For cloud, Terraform, CI/CD, GitOps, SRE, Linux and AI infrastructure we build the hands-on lab around your stack, and every area is also covered in expert-led interviews.

How is the score calculated?

Each lab has a set of automated checks that inspect the final state of the environment, each worth a number of points. A reviewer can add qualitative comments on approach and method. The score is one input to your decision, not a verdict.

Can candidates use documentation or AI tools?

You decide the rules for each assessment. Labs are designed so that the outcome depends on investigating a specific environment and validating the fix, which rewards real understanding regardless of tools. We do not claim to detect every use of AI assistance.

Can you customise a lab for our team?

Yes. We can adapt any ready-made lab or build a new one around your cloud, tooling, conventions and the incident types your team deals with. Tell us the role and what you want to test; custom lab work is scoped and quoted per requirement.

What data do you keep about candidates?

Only what is needed to produce the assessment report: the candidate's name or identifier, the lab result, checks and command history. Retention periods are agreed with each client. This marketing website does not collect candidate assessment data.

See a lab running on a real cluster.

We will walk you through a live assessment and suggest labs for your open roles.

Request a demo Discuss requirements

Hire the engineer who can actually fix it.

Better signal, with less of your engineers' time

See real working method

Compare consistently

Save senior time

Decide on evidence

How an assessment runs

Define the role

Invite the candidate

Candidate works the problem

Review the results

Matched to the seniority you are hiring for

Example lab scenarios

CrashLoopBackOff after a release

Pod stuck in Pending

ImagePullBackOff and a stalled rollout

CreateContainerConfigError

Service cannot reach its Pods

Cluster DNS is down

RBAC permission denied

Liveness probe kills a slow starter

PersistentVolumeClaim stuck in Pending

Bad release and a stuck rollout

Node maintenance gone wrong

NetworkPolicy outage

Broken admission webhook

Pod Security Admission rejection

CoreDNS crashes after a config change

ResourceQuota deadlock

New Services do not work

Worker node NotReady

kubelet will not start

kubelet cannot reach the API server

Container runtime down

DNS broken on one node only

Node under false disk pressure

API server down

etcd will not start

etcd database full

Scheduler down

Controller manager down

Survive node maintenance

Multi-tenant isolation

Autoscaling that will not scale

Enable API audit logging

Encrypt Secrets at rest

Failing GitHub Actions pipeline

Image build cannot push

Canary release gone wrong

Argo CD application stuck OutOfSync

Flux reconciliation failure

Helm upgrade fails mid-release

Terraform state drift and lock

Refactor modules without replacement

Cloud IAM access denied

The alert that never fires

Define SLOs and burn-rate alerts

Production latency investigation

Disk full on a busy host

systemd service will not start

Intermittent connection resets

Enforce signed images

Secrets exposed in workloads

GPU node not schedulable

LLM inference server keeps restarting

Inference platform design review