<![CDATA[We present a face-recognition-based employee attendance system built on a microservices architecture and integrated with the external Worker AI (LSKK) inference API. The design separates camera I/O, verification, persistence, and presentation into independently deployable services, enabling targeted scaling and resilient operation through asynchronous queues. The system was developed using Rapid Application Development (RAD) and evaluated via black-box testing that covered authentication, camera and AI data views, filtering and pagination, reporting, and employee CRUD. The results show conformance to specifications: the interface renders the expected outputs, and the cooldown policy effectively prevents duplicate entries, while the separation of history (raw) and history_ai (verified) supports traceability and clean reporting. These findings indicate that combining microservices with API-based face recognition offers a practical and maintainable alternative to RFID-based workflows with fewer operational frictions. Limitations include the use of an external inference API (model configuration and thresholds are outside our control) and testing within a single organizational setting. Future work will focus on operational measurements of the deployed pipeline, particularly end-to-end latency under load spikes and queue formation, as well as monitoring misread/error rates to inform model improvements.]]>
Copyrights © 2025
