Designed a Siamese Neural Network with shared-weight twin branches and Contrastive Loss to extract discriminative face embeddings for one-shot identity verification and matching., Integrated MTCNN for robust face detection and cropping; developed dynamic positive/negative pair generation with online hard-negative mining to accelerate metric learning convergence., Built a complete pipeline including embedding database management, real-time face matching, and visualization utilities for evaluating verification accuracy on unseen identities.