The product
The interview happens. You just stop going blank.
WhileHired has one job: turn a question you just heard into something you can actually say, without sounding like you are reading. This page is the whole path, including the parts that fail.
Before the call
A few minutes of setup, once.
The Brief is the only thing you have to think about. It is also the thing that keeps the answers from sounding like everybody else's.
-
Add your resume and the job
Upload the resume you actually sent. Paste the job description you actually applied to.
Five minutes, once. This Brief is what every answer is built from. No generic answer bank. No scraped question dump. No credential to keep from a previous role you did not hold.
-
Join the call like normal
Teams, Meet, Zoom, or a phone on speaker. No bot joins, no link to share, nothing changes for the interviewer.
WhileHired listens to the audio already playing on your laptop. The other side installs nothing and sees nothing different. You still take the call.
-
Read one line, then speak
A twelve-word glance line lands about a second after the question ends. Three beats wait behind it.
You speak in your own words. The Cue is a prompt, not a script. If your resume does not support a claim, it says so instead of inventing one.
-
Review it after
The transcript, the questions, and the suggested answers wait in Recall. Or delete the Thread on Stop.
One switch decides whether the interview is kept. Turn it off and the Thread is gone, not archived quietly for later. You decide what is remembered.
Resume
PDF, DOCX, Markdown or plain text. Size and type are checked server-side, not in the browser.
Job description
Paste the text. One active role at a time keeps the Brief honest about which team you are facing.
Your voice
Answers are prompts, not a script. You still have to say the thing — which is why it sounds like you.
During the call
One line to read. The rest is still you.
A glance line lands about a second after the question ends. Three beats sit behind it if you need more. You speak in your own words.
| State | What it means | Colour |
|---|---|---|
| Listening | Microphone and system audio are live. The transcript fills in as people speak. | live |
| Thinking | A question has been detected and the answer is being assembled. | accent |
| Answering | A glance line and up to three beats are on screen. | accent |
| Paused | Nothing is being captured. The last answer stays visible, marked as paused. | muted |
| Reconnecting | The network dropped. The transcript is buffered locally and will resume. | warn |
| Error | Something failed and the Cue says what. Retry, or stop the Thread. | danger |
Interviewer asked
“Tell me about a time you cut latency on a system that was already in production.”
Cut p99 latency 42% by replacing a synchronous fan-out with a queued pipeline.
- Traced the tail to one downstream call, not the database everyone blamed.
- Shipped behind a flag, ran both paths for a week, compared real percentiles.
- Wrote the runbook the on-call team still uses for queue depth alerts.
Grounded in your resume · Senior Backend Engineer, 2022–2024. No supporting fact for the migration cost, so it is not claimed.
Transcript 6 / 6
- Interviewer Before we get to the code, walk me through your last role.
- You Three years on the payments platform, mostly on the ledger service.
- Interviewer Tell me about a time you cut latency on a system that was already in production.
- Listening and how did you convince the team…
What the interviewer sees
Your shared screen, your camera and your voice. The Cue window is excluded from screen capture, so it is not in the share and not in their recording.
What the interviewer hears
You, answering. WhileHired never speaks, never joins, and never sends a message to the other side.
When it breaks
Four failures we designed for, because they will happen.
A realtime tool is judged by what it does when the connection is bad and the room is quiet. None of these paths fail silently.
- The network drops
- The Cue says Reconnecting, backs off up to 30 seconds, and keeps buffering what it hears. When the connection returns, the transcript continues without duplicate lines and without re-answering a question you have already seen.
- The microphone is denied
- It does not pretend to listen. The status bar shows the failed source, and the Cue tells you exactly where to enable it in your system sound settings.
- The AI provider times out
- You get an error card with a retry, not a spinner that never resolves. The Thread keeps running so a single slow answer does not end the interview.
- The app crashes
- It recreates the window once and shows a toast. A second crash inside a minute ends the Thread cleanly and finalises usage exactly once — no double charge.
Privacy
What leaves your machine, and what does not.
Audio is streamed to the transcription provider with a short-lived, single-use token scoped to your Thread. It is converted to text and dropped. The provider does not receive your resume; the language model does not receive your audio.
Transcript text and the answers shown are stored only when Recall is on. When it is off, deleting the Thread deletes the text. No transcript body ever appears in our logs.
-
Never leaves
Your provider keys, your prompt templates, and any way to raise your own limits. They are not in the binary.
-
Leaves, then is discarded
Audio frames, for as long as it takes to transcribe them.
-
Stored, if you allow it
Transcript text, detected questions and suggestions, readable only by your account.
Start free
Bring one real question to it before the call.
The honest test is your own resume and a question you were not expecting. Thirty minutes are free, then ₹1,079 for 3 hours, GST included.