Automated attendance systems have become an integral component of modern smart campus infrastructures. However, most existing solutions rely on centralized process- ing pipelines that introduce high latency, network dependency, and significant privacy risks due to continuous transmission of sensitive biometric data. To address these challenges, this paper proposes a distributed edge-AI framework for real-time attendance monitoring, where facial recognition is performed locally at the classroom level and only essential metadata is synchronized with a central server.

The proposed system introduces a hierarchical recognition strategy that prioritizes staff identification over student matching, thereby reducing cross-identity misclassification during simulta- neous entries. Furthermore, a natural language-driven analytics interface is integrated to enable non-technical administrators to query attendance records conversationally. A real-time prototype is implemented using open-source tools and evaluated under practical conditions. Experimental results demonstrate low- latency recognition, reliable identity segregation, and improved usability.

The proposed framework bridges the gap between intelligent edge computing and real-world campus management, offering a scalable, privacy-aware, and interactive attendance solution