First-round technical interviews, with the evidence for every claim.

Sophia interviews each candidate against the skills you defined, quotes the answer behind every score, and tells you which skills it could not assess, instead of quietly scoring them anyway.

In pilot now with the recruiting team at Highbrow Technologies.

Skills assessment2 of 3 assessed
React state managementStrong

“I'd lift it into a reducer once three components need it. Context alone re-renders everything below.”

API designAdequate

“Versioning in the path. I've not had to deprecate one yet.”

Postgres indexingInsufficient evidence

Not covered before time ran out. No score inferred.

Postgres indexing was marked must-have and the interview did not reach it. The result is incomplete, and says so.

Sample report · see a full one

From a job description to a scored interview.

Set it up once and every candidate for that role gets the same interview.

  1. 01

    Define the role

    Add the job description and the skills you want tested. Mark the ones a candidate cannot do without.

  2. 02

    Invite your candidates

    Choose a window of a week or two. Each candidate gets a private link and takes the interview whenever suits them.

  3. 03

    Read the report

    Every skill scored with the answer that justifies it, and every skill the interview never reached, named.

A report that admits what it missed.

Most AI interviewers hand back a confident score for every skill you listed, whether or not the conversation ever got there. Sophia scores what it covered, quotes the answer behind each one, and names what it did not reach.

Senior Frontend Engineer

Candidate interviewed 12 Sep · 21 min 14 s · English

SampleIncomplete: 1 must-have unassessed

Must-have skills

React state managementStrong
I'd lift it into a reducer once three components need it. Context alone re-renders everything below, and you notice that the moment the tree gets deep.
TypeScriptStrong
Generics on the response type, so the caller gets the shape back instead of unknown. I avoid enums, because a union of string literals does the same thing without the runtime object.
Postgres indexingUnassessed

The conversation ran out of time before this was reached. No score has been inferred, and none should be read into its absence.

Nice to have

CI/CD · AdequateAccessibility · StrongGraphQL · Unassessed

What to do next

Two of three must-have skills assessed, both strong. Before making a decision, cover Postgres indexing yourself in the next round.

Sophia recommends; it never shortlists or rejects anyone. You make the call.

Interview integrity

Clean

Candidate visible for 99.9% of the recording. Two moments logged for review, both consistent with ordinary thinking.

Also included

  • Full transcript, timestamped
  • Interview recording
  • Your private reviewer notes
  • PDF download

What the candidate goes through.

Screening tools fail when candidates abandon them. Sophia asks almost nothing of the person being interviewed.

No account to create
A private link in an email, and nothing to remember.
Whenever suits them
Evenings, weekends, any time inside your window. Nobody coordinates calendars.
Up to thirty minutes, on a laptop
You choose the length. It runs in the browser and needs a camera and microphone, so a computer rather than a phone.
Nothing hidden from them
They are told they are speaking to an AI interviewer, and asked to consent to the recording before anything starts.

Evidence you can check, not a cheating score.

Sophia reviews the recording after the interview and gives you a timeline of anything worth a second look. A glance at a phone, a second face, time away from the camera. Each one is a timestamp you can click and watch for yourself.

It never accuses anyone and it never moves the candidate's score. You see what happened, and you decide whether it matters.

Clean

No second face, no time away from the camera, and no sign of anything being read off screen.

Candidate on camera for the whole interview

03:25
Looked upward while thinking

11:40
Looked upward while thinking

Credits are on us during the pilot.

Pilot workspaces get their interview credits from us, and payments are switched off until the pilot ends. Pricing after the pilot is not final yet, so it is not on this page.

  1. 01

    Reserved when you schedule

    One credit is one minute of interview time. Scheduling a 20-minute interview sets aside 20 credits, so you never send an invitation your balance cannot cover.

  2. 02

    Charged when it is done

    When the interview ends, you are charged for the minutes it actually ran, and never more than was reserved. The rest goes back to your balance.

  3. 03

    Released if unused

    If a candidate never takes the interview, the reserved credits come back when the invitation expires. Cancelling a scheduled interview returns them too.

What Sophia doesn't do.

You will find this out in week one anyway. Better you read it here.

Interviews are in English.
Other languages are not supported yet.
No ATS integration yet.
Today you download each report as a PDF.
Sophia never decides who to hire.
The report gives a recommendation, but there is no automatic pass or reject, and candidates are never shown a score. You read the evidence and make the call. That one is on purpose.

Run your first interview this week.

Sophia is in a private pilot. We set up each pilot workspace and add its credits. If you've been invited, sign in and schedule your first candidates.