Confidential (recruitment) · Recruitment
An end-to-end AI interview platform. Voice and transcription run in one real-time session, with no manual steps afterwards.
Category
AI
Year
2025
Industry
Recruitment
Optimization
Nothing done twice
How it is optimized
Nothing done twice
Voice and transcription share one real-time session.
The interview and its transcript run within the same OpenAI Realtime session, not as two separate passes over the audio. Meeting links, invitations and the transcript email follow automatically, leaving zero manual steps after a session ends.
What keeps it lean
- 01One session for voice and transcript
- 02No second pass over the audio
- 03Follow-up automated end to end
Caching, batching and reuse instead of regenerating the same work.
The challenge
Running structured interviews at scale, whether general, podcast-style or for recruitment, meant someone scheduling every session, someone facilitating it live and someone handling the transcript afterwards. The format was repeatable; the effort around it was not. The client needed interviews that ran themselves from invitation to transcript.
The work
We built a platform where the user chooses an interview type, description, style and a custom intro or outro. It conducts the interview live through the OpenAI Realtime API with the transcript produced in the same session, generates meeting links and invitations, supports several candidates and CVs in recruitment rounds, and emails the transcript the moment the session ends.
Inside the product
More than one screen.
3 more screens from Realtime Voice Interview Assistant, captured at device size.
Candidates
Every round scored on the same rubric, with live sessions one click away.
Interview report
Rubric scores, timestamped highlights and a recommendation to review.
Role setup
Question structure, adaptive follow-ups and voice settings per role.
On the phone
Home
Candidates
Interview report
Role setup
How it is built
Architecture
Next.js powers both the frontend and backend, keeping interview setup, invitations and post-session delivery in one application. The OpenAI Realtime API runs the live voice conversation and transcription, including recruitment sessions with multiple candidates and CVs.
Key features
Challenges & solutions
Challenge
A live interview had to keep the spoken conversation and transcript aligned throughout the session.
Solution
Used the OpenAI Realtime API for both the voice interview and live transcription so they run within the same real-time session.
Challenge
Recruitment interviews needed to support several candidates and CVs without mixing their context.
Solution
Structured recruitment sessions around candidate-specific inputs while keeping scheduling, invitations and transcript delivery in one workflow.
The impact
What became easier, better or more effective.
Interviews that needed a scheduler, a facilitator and a transcriber now run end to end on their own. The recruiter decides the format and the questions; the platform books, conducts and documents the session, and intern recruitment rounds are scored on the same rubric for every candidate.
30+
interviews completed on the platform
Intern recruitment rounds run live through the AI interviewer, each candidate scored on the same rubric.
3
interview formats
General, podcast-style and recruitment sessions, each with a custom description, style, intro and outro.
0
manual steps after a session ends
The transcript is generated live and emailed automatically when the interview closes, so nobody types up a session.
Stack
Frontend & Backend
- Next.js
AI
- OpenAI Realtime API
Optimization
Nothing done twice
Caching, batching and reuse instead of regenerating the same work.
Conscious AILet's talk
Tell us where you are. We'll say plainly what we would do next, and whether we're the right team for it.