Evident

Evident — an AI interview panel that proves its feedback.

Interview practice today is a chatbot and an arbitrary number. Evident is a live voice panel: two AI interviewers with genuinely different jobs — Priya grades technical depth and trade-offs, Sam challenges customer impact and clarity. They listen to what you actually say, take the floor from each other by rule, interrupt and get interrupted gracefully, quote your own earlier claims back at you when you contradict yourself, and adapt difficulty to how you're doing. The whole interview runs in English, Hindi, or Hinglish.

The architecture is the point: an LLM only labels each answer — pure, unit-tested code decides who speaks next, why, and when the interview ends. Scores are a deterministic function of an evidence ledger, and every quote in the report is verified verbatim against the transcript — feedback that cannot be hallucinated, ending only when all four competencies have real evidence, never on a timer.

Then it closes the loop: your weakest skill with the exact answer that cost you, a written lesson, a targeted quiz, and a focused re-interview — with a progress page that proves the score moved. One Agora agent, per-turn voice switching, a single OpenAI-compatible endpoint as the panel's brain, and an offline test suite that runs the entire Director with every vendor down.