Purpose – This study evaluates a contactless employee-attendance prototype designed to recognize faces using a camera near an entrance and register attendance without requiring employees to stop at a conventional device. Design/methods/approach – The prototype combined Histogram of Oriented Gradients (HOG) for facial feature representation, K-Nearest Neighbor (K-NN) for recognition, and Particle Swarm Optimization (PSO) for K-NN parameter optimization. The training set contained 180 facial images from 20 employees (nine images per employee). Evaluation used one image each from 27 participants, including 20 enrolled employees and seven individuals outside the enrollment set. Findings – Of the 27 test images, 25 produced the expected outcome, corresponding to 92.6% accuracy (TP = 18, FP = 0, FN = 2, TN = 7). Two enrolled employees were classified as unknown, while all seven unregistered participants were rejected. Recognition may be sensitive to illumination, pose, camera angle and image sharpness; the present test did not isolate these effects. Research implications/limitations – The small, single-image-per-participant test limits generalization, and the study does not report a direct K-NN versus PSO-optimized K-NN comparison. The prototype also responds to faces shown in photographs, indicating a presentation-attack vulnerability and a need for separate liveness evaluation. Originality/value – The work investigates the integration of established HOG, K-NN and PSO methods into an entrance-based attendance workflow. Its contribution is a prototype implementation and a bounded account of recognition outcomes and practical limitations rather than evidence of superiority over alternative classifiers or deployment-ready security.