Claude grades the work. You grade Claude.

Cost per call:~₹4 · Claude

Every submission is scored by our frontier evaluation layer against the rubric you defined. Before the AI score is revealed to anyone, you (the employer) are prompted to score the same submission independently. We track the delta — and publish the rolling rate live.

Agreement rate live — first data pending

API endpoint

POST /api/submissions

+ GET /api/evaluations/{id}

Model routing

Frontier evaluation layer

Cost envelope

~₹4 per submission evaluated

How rubric scoring works

1. Employer defines rubric — N dimensions, weighted, with 0-100 descriptors per dimension.
2. Candidate submits milestone work.
3. Our frontier evaluation layer evaluates each dimension, citing specific submission evidence for each score.
4. Weighted aggregate = the ProveIQ Score for that milestone.
5. Employer is prompted (before seeing AI result) to rate the same submission.
6. Delta logged in EmployerScoreRating. Published live at /agreement-rate.
7. Employer can override the AI score with a written reason — override is transparent on the candidate's public profile.

See live agreement rate Post your first rubric