Apna Interview Agent
Key point
Apna leveraged ElevenLabs technology to build a realistic, ultra-low-latency AI interview agent.
Details
India's recruitment platform Apna built an advanced AI interview ecosystem to provide customized mock interviews for over 60 million users and more than 10,000 companies.
This system was implemented by combining ElevenLabs's Text to Speech (TTS) with Blue Machines's voice orchestration platform. Through this, it achieved ultra-low-latency performance where voice playback begins within 150–180ms, and to date has processed over 1.5 million AI interviews and 7.5 million minutes of voice data.
The key reasons for choosing ElevenLabs are as follows:
- Ultra-low-latency streaming: Only 150–180ms of latency occurs until response playback, maintaining the flow of conversation.
- Multilingual support: Seamlessly supports Indian English, Hindi, and code-mixed speech.
- Emotional nuance: Enables tone adjustment that mimics human empathy and a challenging attitude.
Blue Machines operates approximately 500 million micro models by generating a RAG (Retrieval-Augmented Generation) graph for each intersection of role and company. This enables sophisticated interviews that reflect each company's evaluation criteria, tone, and specialized terminology.
The entire process, from recognizing the candidate's voice to generating and playing back questions, is completed within approximately 300ms. This is the threshold at which humans perceive the conversation as uninterrupted, providing both technical precision and emotional depth simultaneously.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.