AI Interviewer
Practice English job interviews with AI. Free, no sign-up.
18 followers
Practice English job interviews with AI. Free, no sign-up.
18 followers
Most interview-prep tools are paid or make you sign up and configure before you can practice anything. AI Interviewer strips that away: upload your CV, and the AI reads it, detects your role and level, and builds a realistic interview from your actual experience — not a generic question bank. You answer by speaking out loud, each question read aloud like a real interview, then get a scored report. No account, no payment, no setup. Free, 5 interviews/day.
how does the speech-to-text handle accents or background noise, and is the report based only on what you said or does it factor in delivery too?
@sabriyeorht
Great questions 🙏
On speech-to-text: by default your answer is transcribed live in the browser via the Web Speech API. If that comes back empty (some browsers/devices don't support it well), we fall back to Whisper on the audio, which handles accents noticeably better. We currently auto-detect the language. Background noise can still affect accuracy, so a quiet environment helps.
On the report: each question is generated with a model answer built from your CV, and scoring compares your response against that model answer — so it's based on the content of what you said and how well it addresses the question, rather than delivery (pace, pronunciation) for now.
If you find it useful, I'd really appreciate a quick review or any feedback — it genuinely helps me improve it 🙏
Reading interview questions or preparing answers in your head is completely different from saying them out loud under pressure. I like that AI Interviewer focuses on that actual moment instead of just giving another generic question bank. the no account, no payment, no setup flow also makes a lot of sense here, because interview prep is exactly the kind of thing people often want to start immediately.
Curious how the scoring handles language vs. substance. if someone has a strong technical answer but weaker English delivery, does the report separate communication feedback from role-specific interview feedback?
@andrasczeizel
Really appreciate this — you've pinpointed exactly the distinction that matters.
Right now, the honest answer is: scoring is based on substance. Each question is generated with a model answer grounded in your CV, and your response is scored against that — so it's really measuring whether the content of your answer addresses the question well, not your English delivery separately. So someone with a strong technical answer but weaker English would currently be scored mostly on the substance.
Separating communication feedback from role-specific feedback is exactly where I want to take it next — that split would make it genuinely useful for non-native speakers, which is who I built this for (I'm one myself). It's the most valuable thing on my roadmap right now.
On the speech-to-text side: I use Whisper, which handles English well across a range of accents. The product currently supports English interviews only — that was my own need first — and non-English CVs are rejected for now.
Thanks for such a sharp question 🙏
How does the AI actually pick which questions to ask from my CV, and does it adapt if I fumble on an early one or just keep following the same script?
@sleymano8vj
Great question 🙏
On picking questions: it works in two steps. First it reads your CV and designs an interview structure — detecting your role and level, then laying out sections (intro, deep-dives into your actual projects, technical, situational). Then it generates the questions for each section, grounded in your real experience rather than a generic bank.
On adapting: right now it follows the planned structure — the full interview is designed up front from your CV, so it won't dynamically branch based on whether you fumbled an earlier answer. It's honest, structured practice rather than a fully adaptive interviewer. Making it adapt in real time to how you're doing is something I'd love to build next — it's a natural next step.
Thanks for digging into how it actually works!
I like that you're removing the setup friction instead of just adding another interview simulator.
As AI becomes better at evaluating people, I think the quality of interview practice will depend less on having more questions and more on how accurately the system adapts to someone's actual background.
That feels like the more durable direction.
@aryan787544
Really well put — that's exactly how I see it too 🙏 The number of questions has never been the bottleneck; it's how well the interview reflects your actual background. That's why I split it into designing the structure from your CV first, then generating questions per section — so it's grounded in your real experience, not a generic bank. And you've named the direction I most want to push: making the system adapt more accurately to each person. Right now it personalizes up front from your CV; adapting in real time to how someone actually responds is the next step. Appreciate you seeing the durable version of this, not just the surface.
@d1019cpu I think that next step could change what users expect from interview preparation altogether. Once the system adapts to how someone actually responds—not just what's written on their CV—it stops feeling like a personalized questionnaire and starts behaving more like a genuine interviewer. That seems like a much harder position for competitors to replicate than simply generating better questions.
@aryan787544
You've articulated it better than I have 🙏 That's exactly the moat — once it adapts to how someone actually responds, not just what's on the CV, it stops being a personalized questionnaire and becomes something that behaves like a real interviewer. Anyone can generate better questions; adapting to a live conversation is a much harder thing to copy.
That's the direction I'm committed to building. Right now I'm running this solo and self-funded, so the pace depends on resources — but this is the core of where I want to take it, and conversations like this make me more sure it's the right bet. Really appreciate you thinking through it with me.
@d1019cpu I appreciate you sharing that. One thought kept coming to mind while reading your reply, but it's a bit too nuanced for a Product Hunt thread.
If you're open to it, send me your email. I'd be happy to write it up—I think it could be useful as you keep shaping this direction.
Heym
Thanks for shipping this, Long. The CV-grounded question generation solves a specific problem for people who can reason through a technical answer silently but have never had to say it out loud in English under time pressure, like a backend engineer who understands a system design tradeoff cold but freezes explaining it in a live interview.
Since the interview structure is planned upfront from the CV, have you thought about letting someone flag mid-session that a question is aimed at the wrong seniority level, instead of only spotting that mismatch after the scored report comes back?
Skipping account creation and payment matters here too, since interview practice is usually something people want to start right before a stressful moment, not after filling out a signup form.
@mbakgun
Thanks, really appreciate the thoughtful feedback 🙏 The mid-session flag idea is a good one — adding it to my list. Cheers!
Tried it with my product manager CV and the questions actually pulled from projects I listed, not the usual generic "tell me about yourself" stuff. The voice reading questions out loud makes it feel way less awkward than typing into a chatbot.
@ekremkodatffu Thanks 🙏 CV-grounded questions and the "out loud" part were the two things I cared most about getting right — really glad it landed.