Evidence-based LLM shortlisting

Built for Ottomater, this service turns a job description and a set of résumés into a ranked shortlist. A large language model extracts explicit requirements and a small set of inferred criteria, enriches them with retrieval, and scores each candidate against each criterion.

Every score comes with a short justification and a longer explanation. The ranking is decision support for a recruiter, not an automated hiring decision. The design includes checks for hallucinated qualifications, biased assessments, and explanations that cannot be audited.

The app is deployed on Google Cloud Run: https://shortlisting-1016040926675.us-central1.run.app