How an assessment works

An Assay assessment has two parts, and a candidate does both in one sitting.

1. The work trial

The candidate is given a real task and a code editor running in an isolated, ephemeral sandbox. They write and run code there, with the tools they would normally use.

The sandbox is provisioned per session and destroyed at the end of it. Candidate code never executes on Kravos infrastructure or inside our application processes.

2. The interview

Afterwards, the candidate is asked about the work they just did. Depending on what the hiring company has configured, this is either a spoken interview with an automated interviewer, or a written one where the candidate types their answers.

The interview is grounded in what actually happened during the trial, not in general knowledge. Someone who did the work can explain their own decisions.

What comes out

A score, its component subtotals, whether it met the threshold the hiring company set, and the written reasoning behind each part, alongside the candidate’s actual work.

Assay does not reject candidates

The result is not a decision. The company that invited the candidate sees the score, the reasoning and the underlying evidence, and decides for itself. A candidate who believes an assessment is wrong can raise it with that company, which can see the full transcript and disregard the result.

What the hiring company configures

• The task, drawn from a set of tickets attached to the role.
• The rubric: named dimensions, each weighted, each optionally with its own threshold, plus an overall pass threshold.
• The interview mode, spoken or written, settable per candidate as an accommodation.