While Artificial Intelligence (AI) and Machine Learning (ML) hold significant promise for Engineering Design Optimization (EDO), traditional optimization approaches frequently suffer from excessive computational and time expenses. To overcome these barriers, this study introduces a high-performance optimization framework built upon three core contributions. First, a cost-effective data acquisition strategy is proposed, utilizing existing manufacturer catalogs alongside data augmentation methods to produce high-quality datasets with minimal resource expenditure. Second, an automated, Genetic Algorithm (GA)-driven approach is designed to optimize the hyperparameters of a Deep Neural Network (DNN), successfully eliminating the reliance on manual expert calibration. Third, a Surrogate-Assisted Genetic Algorithm (SAGA) is deployed, leveraging the highly accurate DNN as a surrogate model to rapidly navigate discrete design spaces and circumvent computationally exhaustive simulations. The practical viability of the framework was rigorously evaluated through an industrial steel grating design application. Empirical outcomes indicate substantial real-world utility, yielding an average mass reduction of 23.21±0.65% across 44 standardized configurations without violating structural or serviceability constraints. The optimization pipeline exhibited remarkable computational efficiency, completing the task in just 3.2 min per model. This acceleration is directly facilitated by the high-fidelity surrogate model, which delivered a classification accuracy of 97.678±0.472% and robust mass prediction metrics (R2=0.999, MAE = 0.719±0.101 kg/m2). Additionally, the methodology demonstrated excellent adaptability by successfully resolving the classical 200-bar truss benchmark subject to strict displacement and frequency constraints. Demonstrating superior performance over existing baseline approaches in both solution quality and execution time, this study substantiates the framework as a highly flexible, scalable, and robust approach for resolving complex engineering design challenges.