An Intelligent Web-Based Multi-Agent Framework for Automated University Timetabling via Constraint-Aware Optimization over an Underlying UniTime Engine

University timetabling is a large-scale constrained optimization problem shaped by dynamic academic requirements, heterogeneous resources, and frequent operational changes. This paper proposes an intelligent web-based scheduling framework that integrates a Multi-Agent System (MAS) supervisory layer with the UniTime solver. The framework encapsulates UniTime within a structured orchestration layer that standardizes constraint modeling, enforces deterministic validation, and supports iterative refinement. Experimental results show higher feasibility, fewer soft-constraint violations, fewer iterations, and lower convergence variance than direct solver deployment. By decoupling user interaction from solver execution, the framework improves deployment robustness and establishes supervisory orchestration as a scalable, institution-ready layer for academic scheduling systems.