High-dimensional shape optimization presents significant computational challenges, especially in contexts that require a faithful representation of physical behavior. Parametric Model Embedding (PME) [1] provides a method for design-space dimensionality reduction [2] by compactly representing key design parameters in a form directly linked to the initial parametrization, simplifying the optimization framework. However, standard PME lacks consideration of the physical aspects of the problem, limiting its relevance for applications that demand physical accuracy. To overcome this limitation, we developed a physics-informed PME (PI-PME) that incorporates low-fidelity physical information, allowing for the creation of reduced spaces that better align with the physical context of the design task. In this work, we demonstrate the application of PI-PME to a bio-inspired autonomous underwater glider (AUG) with a manta-like shape, incorporating low-fidelity physical information to create reduced spaces more closely aligned with the problem's physical requirements. While this study demonstrates the PI-PME methodology on an AUG, the approach is broadly applicable to all shape optimization problems, enhancing optimization frameworks by creating reduced spaces that adhere to both design and physical constraints and making it a valuable tool for high-dimensional shape optimization across various engineering applications. REFERENCES [1] Serani, A., & Diez, M. (2023). Parametric model embedding. Computer Methods in Applied Mechanics and Engineering, 404, 115776. [2] Serani, A., & Diez, M. (2024). A Survey on Design-space Dimensionality Reduction Methods for Shape Optimization. arXiv preprint arXiv:2405.13944.