In the dynamic landscape of social media, understanding of emotional expression and its detection from user-generated content is pivotal for various applications, from marketing strategies to mental health monitoring. Traditional methods of face recognition and emotion detection often fail due to inefficient way of handling complex and unstructured social media posts. This paper proposes a novel architecture, “SocialFaceEmoNet”, that utilizes the power of deep neural networks (DNNs), specifically Convolutional neural Networks (CNNs) and Transformer models to extract features for simultaneous recognition of faces and detection of emotions with high accuracy and efficiency. The proposed architecture begins with the curation of SocioFaceSet, a specialized dataset tailored to the unique characteristics of social media imagery. Unlike previous approaches that rely solely on pre-trained models, SocialFaceEmoNet incorporates transfer learning techniques to adapt the CNNs and Transformers to the nuances of social media data, thereby improving their performance. Furthermore, we introduce a comprehensive evaluation framework to assess the effectiveness of our approach. Through extensive experiments on real-world social media datasets, we demonstrate the superiority of SocialFaceEmoNet over existing methods in terms of both face recognition and emotion detection tasks. Notably, our architecture achieves remarkable results even in challenging scenarios characterized by varying lighting conditions, image resolutions, and facial expressions.