Збережено в:
Бібліографічні деталі
Автори: Mrs. J. Veerendeswari, Mr. Kabilan S S, Mr. Logapriyan A, Mr. Rajesh R
Формат: Recurso digital
Мова:
Опубліковано: Zenodo 2026
Предмети:
Онлайн доступ:https://doi.org/10.5281/zenodo.19255340
Теги: Додати тег
Немає тегів, Будьте першим, хто поставить тег для цього запису!
Зміст:
  • Contemporary academic institutions remain constrained by fragmented information silos, labor-intensive administrative workflows, and inflexible permission structures inherent to legacy Enterprise Resource Planning (ERP) platforms. Although Large Language Models (LLMs) present a compelling opportunity to modernize institutional operations, their deployment within multi-stakeholder educational environments introduces non-trivial risks around data confidentiality and intra-organizational access governance. This paper presents the backend architecture of an LLM-augmented College Management System (CMS) purpose-built for administrative and faculty operations, proposing a principled approach to embedding generative AI within the sensitive boundaries of higher education infrastructure. At the core of the proposed system is AIRA - an Adaptive Intelligent Routing Architecture - a multi-agent AI framework orchestrated beneath a rigorously enforced Role-Based Access Control (RBAC) layer. This architecture automates high-complexity institutional workflows including dynamic academic report generation, attendance analytics, and fee lifecycle management, while ensuring that all AI-mediated database interactions remain strictly bounded by the requesting user's authorization profile. Informed by documented vulnerabilities in production-grade LLM agent deployments, the design deliberately decouples AI routing logic from core transactional database operations, and enforces token-authenticated security contracts at every API boundary. The result is a scalable, role-aware blueprint for LLM augmentation in academic administration — one that advances operational intelligence without compromising the integrity or confidentiality of sensitive institutional data.