Title: Local Gradient Optimization of Modular Entangling Sequences

Abstract: Implementation of logical entangling gates is an important step towards realizing a quantum computer. We use a gradient-based optimization approach to find single-qubit rotations which can be interleaved between applications of a noisy nonlocal gate to dramatically suppress arbitrary logical errors, while steering the evolution operator towards the perfectly entangling subset of SU(4) gates. The modularity of the approach allows for application to any two-qubit system, regardless of the Hamiltonian or details of the experimental implementation. This approach is effective for both quasi-static and time-dependent $1/f^{\alpha}$ noise. We also show how the fidelity of the final operation depends on both the fidelity of the local rotations and the noise strength.