Accurate attendance tracking remains a persistent challenge in rural Indian schools, where paper-based registers often cause delays and inaccuracies. Such shortcomings directly affect government initiatives like the Mid-Day Meal (MDM) scheme, which depends on reliable data for resource allocation. This paper presents AttendifAI, a prototype attendance system designed to function in low-connectivity environments using facial recognition and an offline-first architecture. The system integrates a webcam, a lightweight SQLite database, and a synchronization client that updates a centralized dashboard whenever internet access is available. Preliminary analysis demonstrates that the approach is technically feasible on low-cost hardware and capable of reducing manual errors. By minimizing teacher workload and improving the accuracy of government reporting, AttendifAI has the potential to strengthen monitoring mechanisms and optimize resource distribution in rural education.