Autonomously designed pulses for robust, site-selective control of atomic qubits

Quantum computers based on cold-atom arrays offer long-lived qubits with programmable connectivity, yet the relatively low fidelity of site-selective, single-qubit control limits their programmability. We introduce an artificial-intelligence (AI) framework that overcomes this limitation. Trained on atom-laser dynamics in the presence of atomic motion in optical tweezers, a deep neural network autonomously designs robust composite pulses that increase local control fidelities tenfold while remaining compatible with existing control hardware. We further demonstrate the robustness of these pulses regarding optical aberrations and beam misalignment. This approach establishes AI-trained pulse compilation for high-fidelity qubit control, and can be readily extended to other atomlike platforms, such as trapped ions and solid-state color centers.