Project // 04Status // liveYear // 2026

Intervi AI

Automated interviews with live voice interaction, transcript capture, and AI-generated candidate feedback.
DeepgramGroqSocket.IOBullMQMongoDBAWS S3

The Problem

Live interview media, transcription, analysis and recruiter review have different latency and reliability requirements and cannot share one blocking workflow.

The System

  • React — Candidate & recruiter UI
  • Socket.IO — Live session transport
  • BullMQ — Async processing
  • S3 / MongoDB — Media & records

The Build

  • Live voice interaction and transcript capture
  • Chunked video streaming with retry mechanisms
  • AI-generated candidate feedback
  • Recruiter dashboards and interview templates
  • Real-time proctoring with 9 flags
  • Cloud playback and review workflow

Engineering Notes

Note 001

Challenge. Processing long media without blocking live sessions.

Decision. Separate ingest from asynchronous BullMQ workers.

Tradeoff. Eventual results instead of immediate evaluation.

Result

  • 9 proctor flags
  • Async retry
  • Cloud media