We tell you where you'd actually fail — and how to fix it.

Students spend thousands on coaching and still walk into final rounds not knowing which line of their answer would break the room. The feedback they get is mostly generic — “be more specific”, “use the STAR framework”, “tell a better story”. None of it tells you the verdict the actual interviewer would have given.

InterviewMentor was built to close that gap. The verdict bands are calibrated against real behavioural interviews. The probes follow how a real interviewer chases for evidence. The coaching note quotes the exact moment where the answer weakened, instead of vague advice that's hard to act on.

A voice-first mock interview that pushes back when an answer is shallow, asks the follow-ups a real screen would ask, and produces a report that scores the answer at the indicator level — not just the surface score.

The library underneath is hand-tuned. Each behavioural dimension ships with calibrated indicators, scope tiers, and probe rules, so the verdict the model gives is grounded in something repeatable instead of whatever the LLM felt like that turn.

The product is built for the moment that matters: the night before a real interview, when generic advice is too late and hiring a coach for £200/hour isn't an option.