Quantum computation requires operations that exhibit high fidelity, rapid execution, and resilience to variations in operating conditions. In this work, we combine two approaches to achieve these merits: adiabatic evolution and robust optimal control. Protocols that are based on the adiabatic theorem are inherently robust but are also inherently slow. Many efforts are dedicated to speeding up adiabatic protocols while preserving their robustness. Here, we explore a recent construction provided by the inertial theorem [1] to find rapid protocols that are non-adiabatic in the lab frame but are adiabatic in a specific reference frame. Then, we improve the performance of inertial protocols by using quantum optimal control (QOCT). QOCT has proven highly useful in achieving record fidelities in qubit control [2]. However, as the computational problem becomes bigger, brute force optimization can become impractical. Combining semi-analytic solutions with numerical optimization is a promising path for addressing this challenge. Specifically, we extend the QOCT algorithm of Gradient-Ascent Pulse Engineering (GRAPE) to find control protocols that optimally satisfy the adiabatic theorem. We do so by favoring pulses that maximize the overlap between the simulated state and the instantaneous eigenstate. We derive analytic expressions for the cost functional and its gradient and apply this algorithm to several protocols, including population transfer in two- and three-level atoms (RAP and STIRAP), quantum logic gates, and adiabatic teleportation. We analyze potential realization of our protocols with Rydberg atom arrays. Fig. 1 compares the fidelity and robustness of adiabatic teleportation, implemented using three control pulses, 𝐽 , (𝑡) and 𝐽 , (𝑡), shown in (b). The red pulse is found by using the inertial theorem. Results of our QOCT code and a detailed noise analysis for adiabatic quantum logic gates can be found in our preprint [3].