# Auto Recruiter
a web application that utilizes multi-agent to automates first-round hiring through real-time voice interviews. From a simple needed position detail into complete, unbiased report without having to spend hours on repetitive screening calls. Check the project repo here
Simple Input. Complete Process.
Just provide the position details, and the autonomous agent handles the entire screening pipeline.
## Detail Analysis
The system evaluates the candidate's technical accuracy while strictly grading their soft skills across a four-dimension MECE framework (Clarity, Structure, Assertiveness, and Active Listening), giving hiring managers an unbiased, highly structured an comprehensive, data-driven report breakdown of the candidate's true capabilities.
## Cheating Check
Actively monitors for prompt injection and manipulation attempts using a robust 3-layer architecture. By utilizing the industry-approved deberta-v3-base-injection model from Deepset, it instantly detects and blocks candidates trying to trick or hack the AI into giving them a passing grade, providing enterprise-grade security.
## Real-Time Voice
The core conversational experience is powered by LiveKit's WebRTC and Deepgram to deliver human-like, ultra-low latency audio streaming. Instead of the awkward delays typical of basic AI wrappers, this streaming architecture processes speech-to-text and text-to-speech in milliseconds.
## Agent Tracing & Observability
To ensure enterprise maintainability and reliability, every multi-agent execution step is fully traced with Langsmith. This provides complete end-to-end observability across all LLM chains, tool invocations, and agent state transitions for seamless debugging and performance monitoring.