This study proposes a novel Multi-Agent System (MAS) framework powered by large language models to enhance fairness, transparency, and efficiency in graduate admissions. The MAS architecture features specialized agents—including a Diversity, Equity, and Inclusion Agent and personality-calibrated evaluators—that assess applicants holistically while addressing biases stemming from cognitive distortions, reviewer personality traits, and systemic inconsistencies. It integrates human-AI feedback loops, fairness auditing techniques, and explainable decision-making to promote equitable outcomes. This research advances AI-assisted selection processes by operationalizing fairness principles through agent collaboration and adaptive, auditable decision workflows.