AI Powered Interview System

The Challenge

Past a certain volume, assessing tutor applications by hand is impossible. Scheduling a call, watching the recording, scoring and comparing kept candidates waiting for days and left the judgement shifting from reviewer to reviewer. Video, audio, AI and a hiring workflow had to become one system.

The Approach

1. // 01The application form and the interview system were connected with API keys and scoped permissions; the invitation goes out by SMS and email the moment a candidate is created, behind rate limits and subject level access control.
2. // 02A WebRTC video interview runs in the browser, with recordings uploaded in resumable chunks to secure storage.
3. // 03Speech is transcribed and a GPT-4 based evaluator produces a score, its reasoning, strengths and areas to improve, scaled separately for technical skill, communication and pedagogy.
4. // 04Evaluation state is tracked in its own column (pending, processing, completed, failed) and reflected live to both candidate and admin, with failed runs re-queued automatically.
5. // 05AI hosted briefing videos and a voice assistant layer let the candidate move through the process as if talking to a person.
6. // 06Approved candidates have their profile and payment details pushed into their tutor account, closing the loop from first contact to going live.