AI Interview Platform

Problem

Recruiters needed to run high-volume technical interviews without proportional interviewer headcount — with reliable video capture, scoring, and audit trails.

Architecture approach

Microservices with dedicated video ingestion, transcription, and inference workers. Async pipelines decouple candidate flow from ML scoring latency.

Key challenges

• Video upload and processing at variable network quality
• Isolating ML inference from user-facing API latency
• Scheduling and retry logic across long-running workflows

Scalability decisions

• Queue-based video processing with horizontal workers
• Separate read models for recruiter dashboards vs scoring pipeline
• Idempotent event handlers for at-least-once delivery

Outcome

Platform supports concurrent interview sessions with predictable recruiter turnaround — scoring and playback without blocking the candidate experience.

Tech stack