How vetted assessments are built and scored

There are two tracks. The track follows the skills a talent declares, and the talent can choose it themselves. Neither track runs a multiple-choice skills test today.

• Technical: an AI text interview, then a coding round of two problems generated for the sitting (one of core logic, one real-world task) in a language the talent picks, followed by questions about the code they just wrote.
• Non-technical: an AI text interview about how the talent works, including questions built on realistic workplace scenarios in their role and industry.

The interviewer writes each question from the role, the track, the talent’s experience level, their declared skills and industries, and the conversation so far. It never sees a name, contact details, a photo or a work history, and it never asks for audio or video.

There is no item bank and no psychometric equating. Sittings are kept comparable by the fixed structure, the same rubric and the same weights. We have not yet run statistical equivalence studies between sittings, and we list that below as a limit.

Interview score

The mean of four dimensions: communication, depth of knowledge, problem solving and relevant experience. Each is rated on a five-level rubric, every level adds 25 points, and each rating quotes the answer it rests on.

Coding score

A weighted blend of five dimensions: correctness, approach, code quality, real-world judgement and how well the talent explains their own code in the follow-up answers.

Levels follow experience

Every level is judged against the bar for the talent’s experience level, so a junior and a lead are each measured against what their level should show.

Vetted score

Technical track: AI interview 40% and coding round 60%. Non-technical track: AI interview 100%. The weights are platform defaults and the same for every talent on a track.

How long a score lasts

A vetted score stays valid for six months (180 days) from the day it is issued, and every report shows the date it is valid until.

Job score

On each application, a company ranks by a job score: the vetted score 35%, the job stage (its own interview) 40% and the match with the role 25%, by default. Companies can change these weights per job.

Second opinion and human review

A second AI model re-scores some results: a wide spread between dimensions, a result the scorer flags for a closer look, a score close to a company’s cutoff, an appeal, and a small random sample. Where the two disagree, the disagreement is logged for a person to review. It never changes an outcome by itself.

Scores inform people; they do not decide. A person makes every hiring decision, and nobody is declined automatically. Talent can appeal any score.

During the interview, paste is blocked and focus or tab changes are recorded. During the coding round, pasted code and focus changes are recorded. These signals are advisory: they route a result to a human reviewer and never decide an outcome on their own. Talent can appeal an integrity flag.

The interview is untimed. The coding round has a time limit, and accommodations are available, including extra time, untimed screen-reader sittings and a low-bandwidth mode.

We compare selection rates across self-disclosed demographic groups using the four-fifths rule. Groups below a minimum size are not reported, because ratios from very small groups are not reliable.

This is a monitoring signal, not a certification. An independent bias audit is planned and has not been completed; its status is published on the bias-audit disclosure.

Companies are asked for a 90-day hire-quality check-in on the people they hire, so that whether scores predict performance on the job can be studied in future. No validity results have been published yet.

• There is no published criterion-validity study yet.
• Interview questions and coding problems are AI-generated for each sitting, so content varies between sittings even within the fixed structure, and no statistical equivalence study has been run.
• Submitted code is not yet run against test cases. Correctness is judged by the scoring model reading the code against the task and its examples.
• The interview is text only, so it favours fluent writers.
• Vetted assessments are available in English only.