This repository introduces an approach for the inverse design of metasurfaces using Diffusion Schrödinger Bridges, a powerful generalization of diffusion models. By integrating enhanced posterior sampling techniques, our method overcomes the limitations of traditional ancestral sampling, enabling the generation of high-quality, diverse, and complex metasurface designs. Advanced Posterior Sampling Strategies: We introduce two complementary categories of advanced sampling techniques (Direction and Amplitude-constrained) to enhance the design process. These categories can also be combined synergistically to achieve even more refined results. Direction Posterior Sampling: Posterior Sampling (PS) PS Monte Carlo PS Robust Amplitude-Constrained Posterior Sampling: PS with Spherical Gaussian Constraint PS with Disk Gaussian Constraint PS with Ring Gaussian Constraint These techniques significantly enhance the quality, diversity, and performance of generated metasurface configurations, allowing for the exploration of intricate and high-performance designs. Scalability and Adaptability: The Diffusion Schrödinger Bridge framework supports scaling the inverse design process to larger metasurfaces than those encountered during training. This adaptability makes it highly suitable for real-world applications and outperforms similar posterior sampling techniques used with traditional diffusion models. Robust Training with Consistency Loss: We introduce a consistency loss-based technique during training, which reduces the sensitivity of diffusion models to hyperparameter choices. This innovation improves both the robustness and convergence of the training process, ensuring reliable and efficient design optimization. Custom Noising Schedules: A tailored noising schedules are implemented to optimize the diffusion process, ensuring better control over the generation quality and stability. Applications: This method is particularly valuable for researchers and engineers working on metasurface design, nanophotonics, and electromagnetic optimization, offering a scalable and robust solution for inverse design challenges.