VeriProof: Skill proof portfolio and candidate verification platform

Many technical recruitment methods rely on text resumes and applicant tracking systems (ATS), leaving candidates the opportunity to easily pad their credentials and claim work that is not their own. This paper introduces VeriProof, an automated credentialing and candidate verification platform designed to mitigate the risks associated with AI-generated and exaggerated resumes. VeriProof leverages artificial intelligence to analyze resume claims and cross-reference them with automated GitHub repository intelligence, matching claimed competencies against commit authorship, repository depth, and commit frequency. Furthermore, the platform implements a two-phase adaptive testing engine: Phase 1 calibrates baseline core engineering syntax and concepts, followed by a secure, locked Purgatory Intermission (5–15 minutes) for asynchronous test item synthesis, and Phase 2 delivers tailored, practical debugging scenarios. Test questions generated via LLMs are vetted through a three-pass Question Quality Pipeline enforcing strict mathematical thresholds on distractor length variation and jargon density. Candidate integrity is continuously assessed using non-binary forensic telemetry—tracking window blur and tab-switching events, response latency distributions, confidence calibration divergence, and deceptive trap questions targeting AI hallucinations—synthesizing a holistic Forensic Trust Score (0–100%). Empirical evaluation across 100 simulated trials demonstrates that VeriProof reduces required assessment items from 30 to an average of 18–22 questions (a 33% reduction in assessment fatigue) while preserving discrimination fidelity and effectively isolating fraudulent claims.