Bio-inspired layered structures, particularly Honeycomb and Bouligand architectures, are renowned for their exceptional mechanical properties, such as high strength-to-weight ratios and superior toughness. However, the complexity of their hierarchical designs makes traditional trial-and-error development inefficient and costly. To overcome these challenges, this study presents an intelligent inverse design framework that leverages Artificial Intelligence to precisely map desired mechanical behaviors to specific geometric parameters. Integrating Deep Learning (DL) and Reinforcement Learning (RL), the proposed system employs U-NET for high-fidelity structural slice generation and a Convolutional Neural Network (CNN) as a rapid surrogate model for property prediction. The core design optimization is executed by a Double Deep Q-Network (DDQN) agent. Driven by a reward mechanism calculated from the similarity between target and predicted stress-strain curves, the agent autonomously explores the design space to converge on optimal solutions. The framework was validated through 3D printing and compression testing, demonstrating that this automated approach significantly reduces design iterations while achieving high precision in material customization.