Customer segmentation remains a fundamental challenge in e-commerce, where traditional clustering methods often fail to capture complex, non-linear relationships within high-dimensional customer data. This study proposes a novel unsupervised deep learning framework that integrates behavioral and demographic features to discover nuanced customer segments in e-commerce environments. We employed a hybrid architecture combining deep autoencoders for feature learning with advanced clustering algorithms including Deep Embedded Clustering (DEC) and Variational Autoencoders (VAEs). Using a comprehensive dataset of 8,068 customers with 12 behavioral and demographic attributes, our approach achieved superior clustering performance with a silhouette score of 0.742 compared to traditional K-means (0.523) and hierarchical clustering (0.487). The framework identified six distinct customer segments characterized by unique purchasing patterns, demographic profiles, and engagement behaviors. Notably, the deep learning approach revealed latent behavioral patterns that were previously undetectable, including a high-value, low-frequency customer segment representing 12% of the customer base but contributing 34% of total revenue. The results demonstrate significant improvements in segment interpretability and marketing actionability, with validation showing 23% higher conversion rates when personalized marketing strategies were applied to the discovered segments. This research contributes to the intersection of deep learning and customer analytics, providing a scalable framework for intelligent customer segmentation in digital commerce environments.