Focused specialist rounds, conducted conversationally by AI — and scored with the same evidence-linked rigor as everything else.
Not a quiz. Aria, our AI interviewer, probes adaptively — following the answer where it goes, pushing past anything rehearsed.
A round goes deep on a single skill instead of sampling several shallowly — so the score means something specific.
Follow-ups are generated from the answer given. Rehearsed material runs out fast; real experience keeps going.
No multiple choice, no timed trivia. The candidate explains their reasoning out loud, as they would at work.
Candidates write real code in a live editor while talking through it — and the AI probes the decisions as they're made, not after.
"You reached for a sliding window over a token bucket — what breaks first if traffic is bursty?"
A skill round is not a separate report. It lands in the same scorecard format, under the same honesty rules, and feeds the same debrief.
Each skill score points at the moment in the round that earned it — a timestamp, a quote, a line of code.
A dodge is reported as a dodge. If the round produced no real signal, it says so rather than inventing a number.
Skill rounds sit in the candidate heatmap next to screening and panel rounds, in one hire / maybe / reject view.