Technical Interviews

A verdict on your candidate. By end of day.

A senior engineer in the candidate's stack runs the technical interview. You get back a level call with the exact lines from the transcript that justify it. No score out of ten, no black box.

English and Russian · 45 minutes · report the same working day

EXCERPT · SENIOR BACKEND ENGINEERSenior
Interviewer
Your queue was backing up under load. Walk me through what you actually did.
Candidate
First I checked whether the consumers were slow or just starved. One consumer group was doing a synchronous write per message, so I batched them. Batching costs you per-message acknowledgement, which means a crash replays the batch. We took that, because the writes were idempotent.
Trade-off namedConsequence statedFailure boundary
Same candidate, later
Honestly I have not had to tune consumer count by hand. I would look at lag per partition first and go from there.
Limit acknowledged
3 of 27 signals shownEvery tag links to its timestamp
How it works

Three steps, and none of them are yours.

01

Send the role and the candidate

A job description and a CV is enough. We read the role first and build the interview plan against it, so the questions are about your stack and your seniority bar.

~5 min of your time
02

A senior engineer runs the interview

Forty-five minutes, structured, recorded with the candidate's consent. The interviewer pushes on reasoning rather than recall, and never tells the candidate the outcome.

booked from your calendar
03

You get the report and make the call

A level verdict, the signals behind it, the risks and the gaps. You decide who moves forward. We never discuss the decision with the candidate and never with your client.

same working day
The report

Evidence, not scores.

A number out of ten tells you nothing you can defend to a hiring manager. Every line points back at something the candidate said.

Level

A junior, mid or senior call

Matched against reasoning profiles calibrated on real hiring decisions, not against a percentile of other test-takers.

Signals

Trade-offs, consequences, failure boundaries

The three things that separate levels in practice, each tagged to the exact quote and timestamp it came from.

Risk

Where the story does not line up

When the CV and the room disagree, you see it, with both sides quoted.

Gaps

What they cannot do yet

Named specifically, with a sense of how far off it is. Useful when the answer is “not for this role, but yes for the other one”.

Pricing

Per interview. No seats, no minimum.

You pay for verdicts you actually needed. Compare it against what the same 45 minutes costs when it comes out of your own senior engineer's week.

Single technical interview, report includedon request
Ten or more per monthvolume rate
Recruiters introducing a client10%, see partner program
Get a quote
Questions

The things people ask first.

Who actually runs the interview?+

A working senior engineer in the candidate's stack, following a plan built for your role. Not a recruiter reading a script and not a synthetic voice. You do not see their name and neither does the candidate.

What languages can you interview in?+

English and Russian today. If your candidate is more precise in their own language, use it. Reasoning is what we score, and precision matters more than accent.

How fast is the turnaround?+

The report reaches you the same working day the interview happens. Booking usually lands within two working days of you sending the role.

Do you tell us who to hire?+

No. We tell you what level the person operates at, and what we saw them do and not do. Whether that clears your bar is your call, and the borderline cases are exactly the ones we hand back rather than decide for you.

What happens to the candidate's data?+

Processed for the role you sent it for, under GDPR, and never reused to build a talent pool we sell to someone else. The recording exists because the report cites it. Ask us to delete it and it goes.

We already use an AI screening tool. Why add this?+

Because those tools score answers, and answers are the part a model now produces for free. Reading a submitted artifact is the easier half of the problem. The interview is where a claim gets tested against a person who has to defend it out loud.