For decades, computational simulations have been fundamental for modeling in science and engineering. Researchers and practitioners have relied on simulations for analyzing properties of physical systems, performing trade studies, and use with design and optimization. Forward simulations alone, however, are particularly inefficient for solving inverse and design problems; they require many forward solves because they do not directly provide sensitivity information, such as gradients, that are needed for optimization. Adjoint-based methods provide this gradient information by solving the adjoint equations backwards in time and have been incredibly impactful for engineering and design applications using relatively few queries [1]. While modern autodifferentiation libraries and advances in GPU technology have been key enablers for the modern success of machine learning, these tools have also served as a convenient framework for creating scalable and fully differentiable physics simulators, where gradients are instead computed through backpropagation. Recently these solvers have been used in a variety of domains [2] for challenging applications—such as directly optimizing geometries with respect to design parameters and developing optimal controllers by differentiating through the control objective. These approaches continue to show extraordinary promise as an enabling technology for science and engineering. This minisymposium places an emphasis on (1) recent advances and opportunities in developing differentiable simulations, (2) obtaining gradients for sensitivity analysis of physical systems, and (3) exploiting differentiability to solve challenging control, inverse, and design problems. [1] Giles, Michael B., and Niles A. Pierce. "An introduction to the adjoint approach to design." Flow, turbulence and combustion 65.3 (2000): 393-415. [2] Newbury, R., Collins, J., He, K., Pan, J., Posner, I., Howard, D., & Cosgun, A. (2024). A review of differentiable simulators. IEEE Access.