This study introduces an automated design framework for discovering reprogrammable nonlinear dynamic metamaterials. The framework uses inverse design and differentiable simulations to optimize the geometry of flexible mechanical metamaterials for desired nonlinear dynamic responses. The approach enables the design of materials that can perform tasks such as energy focusing, energy splitting, dynamic protection, and nonlinear motion conversion. The framework is extended to create architectures capable of switching between different dynamic tasks, such as energy focusing and dynamic protection, by utilizing static pre-compression. The designs are physically realized and experimentally tested, demonstrating the robustness of the engineered tasks. The study highlights the potential of the framework in discovering unprecedented material responses through automated design. The framework is capable of handling complex nonlinear dynamics and allows for the reprogrammability of metamaterials, enabling the encoding and switching between multiple tasks. The results show that the framework can efficiently explore the large space of non-periodic architectures and converge to a performant design that encodes one or multiple tasks. The study also demonstrates the effectiveness of the framework in achieving specific dynamic tasks, such as energy focusing and reprogramming the focusing location. The results indicate that the framework can identify optimal designs that exhibit diverse performance trade-offs in focusing at different target locations. Additionally, the framework is shown to be capable of achieving antagonistic tasks, such as maximizing and minimizing kinetic energy at a target location under different pre-compression levels. The study concludes that the proposed framework holds promise in identifying material structures capable of complex transient and steady-state dynamic behaviors in response to simple actuation inputs. The reprogrammability of such behaviors can be further enhanced by enabling simple task selection strategies through pre-deformation, changes in excitation frequency, or variations in loading location. The results demonstrate the potential of the framework in generating soft material embodiments with reconfigurable functionalities.This study introduces an automated design framework for discovering reprogrammable nonlinear dynamic metamaterials. The framework uses inverse design and differentiable simulations to optimize the geometry of flexible mechanical metamaterials for desired nonlinear dynamic responses. The approach enables the design of materials that can perform tasks such as energy focusing, energy splitting, dynamic protection, and nonlinear motion conversion. The framework is extended to create architectures capable of switching between different dynamic tasks, such as energy focusing and dynamic protection, by utilizing static pre-compression. The designs are physically realized and experimentally tested, demonstrating the robustness of the engineered tasks. The study highlights the potential of the framework in discovering unprecedented material responses through automated design. The framework is capable of handling complex nonlinear dynamics and allows for the reprogrammability of metamaterials, enabling the encoding and switching between multiple tasks. The results show that the framework can efficiently explore the large space of non-periodic architectures and converge to a performant design that encodes one or multiple tasks. The study also demonstrates the effectiveness of the framework in achieving specific dynamic tasks, such as energy focusing and reprogramming the focusing location. The results indicate that the framework can identify optimal designs that exhibit diverse performance trade-offs in focusing at different target locations. Additionally, the framework is shown to be capable of achieving antagonistic tasks, such as maximizing and minimizing kinetic energy at a target location under different pre-compression levels. The study concludes that the proposed framework holds promise in identifying material structures capable of complex transient and steady-state dynamic behaviors in response to simple actuation inputs. The reprogrammability of such behaviors can be further enhanced by enabling simple task selection strategies through pre-deformation, changes in excitation frequency, or variations in loading location. The results demonstrate the potential of the framework in generating soft material embodiments with reconfigurable functionalities.