- Voice AI
- Adoption
- EdTech
- 2026
DevMentor
An AI mock interviewer for fresh CS graduates — a real technical interview by voice, then a scored report on how it went.
Situation
Fresh graduates fail interviews they could pass, because the only place to practise is a real one. There is nowhere low-stakes to be bad first, and the feedback from a rejection is silence.
Task
- Give a graduate a realistic technical interview they can fail privately.
- Send them away knowing what to fix, not just how it felt.
Result
Reached 500+ active users within 14 days. A voice interview with a live transcript, then a scored report — strengths, gaps, and a breakdown by skill.
My Contribution
- Made users bring their own API key. That one decision is why 500 people could arrive in a fortnight without an inference bill — the product scales at zero marginal cost, and I never hold anyone's credentials.
- Insisted the output be a verdict, not a transcript. A chat log tells a graduate nothing; a 0-100 score with strengths, gaps and per-skill feedback tells them what to practise on Monday.
- Chose voice over text, because what graduates actually fail is speaking under pressure — not writing answers with time to think.
- Watched onboarding and iterated on the feature set against real drop-off rather than opinion.