QUORUM — Coordinated AI Interview Panel
Link to open source: https://github.com/PrinceMridul/QUORUM
QUORUM is an adaptive, voice-based AI interview panel that evaluates candidates through coordinated Technical, Product, and Hiring Manager perspectives.
Instead of following a fixed question list, QUORUM treats an interview as an information-gathering problem. After every candidate response, it extracts evidence, updates a shared candidate state, identifies what remains uncertain, and selects the highest-value combination of action, topic, and interviewer role for the next turn.
The system can probe vague claims, detect contradictions, adapt difficulty and questioning strategy, switch interviewer roles dynamically, and terminate when sufficient evidence has been collected. At the end, verified evidence is converted into a structured assessment with competency scores, confidence, unresolved areas, follow-up questions, and a final recommendation.
QUORUM is built around a deliberate separation of understanding, state, decision-making, and evaluation. Agora provides the real-time voice interaction, while the QUORUM intelligence engine determines what the panel should learn next and why.
Core pipeline:
Candidate Answer → Evidence Analysis → Candidate State → Adaptive Policy → Role + Topic + Action Selection → Targeted Question → Evaluation
The goal is not to build an AI that simply asks interview questions, but an AI panel that knows what it still needs to learn before making a decision.
This build was uploaded as a hackathon project
