IoT networks are vulnerable to Distributed Denial of Service (DDoS) attacks, which can cause network instability and compromised data integrity. In order to improve detection efficiency and accuracy, this research suggests a revolutionary deep ensemble learning architecture that incorporates pruning strategies. The system uses a Voting Classifier that combines K-Nearest Neighbours, Decision Trees, and Logistic Regression, and a Stacking Classifier that combines Random Forest, Gradient Boosting, and Naïve Bayes models. The machine learning pipeline is automated and optimized by a TPOT Classifier, and redundant models are eliminated for computational efficiency through pruning. The system efficiently differentiates between typical and malicious traffic patterns by using an extensive dataset of network flow characteristics, including packet lengths, inter-arrival periods, and flag counts. The model's exceptional performance is demonstrated by evaluation criteria such as accuracy, precision, and recall, which show that it achieved over 98% detection accuracy with less false positives. IoT network vulnerabilities are addressed by this resource-efficient and scalable method, which provides a strong real-time threat detection solution.