AI Voice Interview Portal
An end-to-end AI voice interview platform for high-volume recruiting. Candidates complete guided voice sessions; audio is transcribed in real time and evaluated against custom rubrics.
Speech Pipeline & Real-Time Rubric Scoring
Project Overview
High-volume recruiting requires screening hundreds of applicants, which typically overwhelms HR teams. The AI Voice Interview Portal solves this bottleneck by conducting guided screening calls asynchronously in-browser.
Our studio shipped the web platform (Next.js), hook-ups for client-side audio streaming via WebSockets, and automatic translation/transcription services using Whisper models. Audio is transcribed and evaluated against customizable rubrics, presenting recruiters with auditable LLM evaluation scores.
Key Deliverables
- Guided real-time audio recording in-browser utilizing secure WebAudio layers.
- Real-time audio streaming and speech-to-text transcription via WebSocket tunnels.
- Bespoke candidate evaluation using LLM-backed rubrics with clear audit traces.
- Comprehensive recruiter dashboards to track pipelines, review transcripts, and filter scores.
- Resilient audio socket session recovery protecting data against user connection drops.
Technologies Used
Client Value
The Challenge
Transmitting raw audio streams in real time from diverse browsers without package drops. Re-establishing the active session context gracefully if the candidate’s network connection was temporarily disconnected mid-interview.
The Solution
Created a custom WebAudio chunking engine streaming 500ms audio buffers over WebSocket lines. Configured a Redis-backed session state layer that auto-buffers audio locally in IndexedDB, syncing pending chunks upon connection recovery.
Measurable Results
Interested in Similar Capabilities?
Discuss speech processing pipelines, WebSocket streaming, LLM-based evaluation metrics, or recruiter workflows.