Interview catalogue
Backend, Frontend, Fullstack across four levels and two languages. Ready to use, editable to fit you.
Voice screening for technical roles
A voice AI interviewer asks your questions, listens to the answers and sends you a transcript. Strictly on your script — no improvisation, no skipped questions, nothing invented.
Russian and English · Backend, Frontend, Fullstack · Trainee to Senior
A 45-minute first interview plus prep and notes. Across a funnel of 30 people that is a mid-level engineer-week pulled out of the sprint.
Two interviewers means two sets of questions and two different bars. In the end there is nothing to compare candidates on.
Booking a slot stretches the funnel by days, and strong candidates take another offer in the meantime.
Pick a ready set from the catalogue — track, level, language — or add your own questions with evaluation criteria and the order you want.
The candidate opens it in a browser, checks the microphone and starts. Nothing to install, no scheduling.
After the conversation you get the full transcript with timestamps, replies tied to questions, and the recorded answers.
How this differs from “wiring GPT to a microphone”
The language model does not know your questions. It physically cannot see them: the list, the expected answers and the criteria never leave the server. All the model can do is ask the backend what to do next — and the backend is free to say no.
A state machine on the server holds the order. The model does not decide what has happened or what comes next — it gets exactly one instruction at a time.
If the candidate said nothing or answered in one word, the attempt to move on is refused and the interviewer asks them to expand.
A question about senior salaries — or anything outside the interview — gets a polite return to the current question. The question itself does not change.
The candidate gets no expected answers, no criteria, no upcoming questions — that data simply is not in their browser.
Backend, Frontend, Fullstack across four levels and two languages. Ready to use, editable to fit you.
Every line is tied to a time and to the question being answered. You can see what was said and when.
Audio is kept next to the transcript: a disputed answer can be listened to rather than argued about. With a retention period and a notice to the candidate.
Wording, topic, difficulty, order, and the list of concepts a good answer should cover.
The interview ends on a server-side timer, even if the candidate closed the tab.
An invite with a limited lifetime that burns on first use. It cannot be forwarded to a friend.
Free
Try it on your own role
€9 / mo
For regular hiring
Billed monthly, cancel any time. The free plan needs no card.
No. The model never receives the question list — it stays on the server, and the model gets exactly one instruction at a time. It cannot skip, repeat or invent a question because it does not know what they are.
The state lives on the server. After reconnecting, the interview continues from the current question rather than starting over.
No. Only the current step reaches the browser. Expected answers and evaluation criteria never leave the server under any circumstances.
Russian and English. The language is set per interview and drives both the wording of the questions and the language the interviewer speaks.
Only if you turn recording on. The candidate then sees a notice before the interview starts, and recordings are deleted automatically once your retention period passes.
No, it replaces the first-round screen. The point is that only the people worth talking to reach your engineer.
We will run it on your questions and tune the script to your process.
Request a demoor email us at hello@interview4you.com