Early access

Hands-on engineering assessments, with AI allowed.

Candidates fix a real broken system with Claude Code or Codex. You see who understands the system and who trusts the AI blindly.

See how it works

Real AI tools in a real stack

  • Claude Code
  • Codex
  • Kubernetes
  • Docker
  • Linux
  • Terraform
  • Git
  • Python
  • PostgreSQL
  • Grafana
A mockup of the Proova dashboard for the example workspace Kestrel Robotics: credits, roles, tasks, invitations by status, decisions, and the invitations of each role.

You bring the candidates. We run the rest.

Proova runs the machines, the invites, the assessment and the grading. You open a ranked list and the evidence behind every rank.

  • InfrastructureA fresh VM for every candidate
  • AutomationInvites, reminders and deadlines
  • AssessmentA real broken system, AI allowed
  • EvaluationGraded against your rubric
A mockup of the Proova candidates page for one role: six candidates ranked by result, each with the invitation status, the score, the evidence, the hiring decision and a link to the report.

How it works

From the role you create to the decision you make.

invite
submit
yes
no
report
You
Create a roleTask, rubric, AI budget
Candidate
Fix a real systemFresh VM · Claude Code or Codex
Understands the system?
Verified
Engineers with AIReads the logs, tests the fix, explains why
Unverified
Trusts AI blindlyShips code it never read or tested
You
You decideEvery prompt, command and test