Chidvi Pemula, founder and CEO of Mirror AI
Companies have recorded and reviewed job interviews for years, studying the footage for hiring decisions long after a candidate leaves the room. Candidates have never had the same option. Chidvi Pemula, founder and CEO of interview preparation company Mirror AI, kept returning to one question inside his own team. If a company can record and study an interview, there is no clear reason a candidate should be shut out of doing the same with their own performance.
That question led Mirror AI to build a feature that lets candidates record and review their own interviews, turning a conversation that used to disappear the moment it ended into material they can learn from afterward. The idea grew out of Mirror AI’s broader mission, helping people get better with every session they run, extended to the one session most candidates never get to revisit at all, the real interview itself. Pemula noticed that candidates spend hours training for an interview and then get almost no chance to study the real thing once it happens, unlike a practice session on Mirror AI, which they can replay and dissect as often as they want.
A Conversation Candidates Already Own
Pemula points to the shape of a typical interview to make his case. Candidates do the vast majority of the talking, describing their own experience, knowledge, and reasoning, while the interviewer asks questions and listens. By Pemula’s estimate, candidates account for roughly 95% of the conversation, with the interviewer’s share closer to 5%. That share includes a candidate walking through past projects, explaining technical decisions, or describing how they handled a difficult situation at a previous job, the exact material a candidate most needs to see played back clearly.
Many of the questions an interviewer asks are already posted publicly on sites like Glassdoor, where candidates share what they were asked after the fact, so the exchange rarely touches anything confidential about the company. A candidate reviewing that conversation afterward is mostly reviewing themselves, their own pacing, their own reasoning, their own answer to a question that was never a secret to begin with. Pemula struggles to find a good reason that a candidate should be blocked from reviewing a conversation built almost entirely from their own words.
Building It Responsibly
Building the feature meant working through real concerns first. Pemula’s team debated the legality of recording an interview. They weighed the privacy of everyone involved, including the interviewer. They also asked a harder question: whether a tool like this would land too close to software built to feed candidates answers during a live interview, a category Mirror AI wanted no part of. They settled on a clear distinction. Helping someone understand how they performed after a conversation has ended is a different kind of tool than one that secretly supplies answers while the interview is still happening.
Recording laws differ from place to place, and the team treated that variation as a core design constraint, building consent into the feature from the very first version. That same thinking shaped a second decision, choosing audio over video. An audio recording captures a candidate’s voice and words without capturing anyone’s image, protecting the interviewer’s privacy by design rather than as an afterthought.
Pemula is clear that the goal was never to collect company secrets or pass along interview questions to other candidates. The feature exists to turn a candidate’s own answers into feedback and targeted practice, much like an athlete reviews game tape after a match to assess their performance. Recording a conversation is nothing new in itself. Candidates already carry phones capable of recording calls, and tools like Granola, Notion, and Zoom transcribe meetings by default. None of those tools analyze what was said or connect one session to the next. Mirror AI’s feature does both, turning a recording that would otherwise go unused into feedback a candidate can act on and build upon with every interview that follows.
An Uneven Starting Point
A basic imbalance in how recording works today reinforces his point. Employers routinely build recording consent into their own interview and application processes, collecting footage they can review and share internally whenever they choose, often without a candidate giving it a second thought.
Data on how many companies record interviews is limited, but a Harris Poll survey cited in a recent analysis from outplacement firm Challenger, Gray & Christmas found that 64% of companies were using video interview tools, platforms that commonly record and store sessions by default. Candidates have no realistic equivalent. Asking an interviewer directly for permission to record, or raising the request during the application itself, risks looking difficult before the conversation has even started, so most candidates simply never ask, and the interview ends up living only in memory.
Pemula wants a legitimate, practical path for candidates to review their own performance, one that does not depend on a request most people are too afraid to make and does not require a company’s permission to exist. Companies have learned from interviews for as long as hiring has existed, building a lasting advantage out of footage candidates never see. Pemula wants candidates to have the same chance, built around their own words rather than someone else’s notes, so that the only side learning from an interview is no longer just the one doing the hiring.