Autonomous closed-loop initial tuneup of superconducting qubit drive and readout

The operation of quantum processors requires reliable qubit tuneup, yet currently employed methods remain largely manual, relying on good initial guesses and extensive error handling.

We present a fully autonomous, closed-loop protocol for first-step qubit tuneup starting from typical experimental uncertainties. Without prior input, the protocol adaptively samples drive and readout parameters using time-domain Rabi oscillations to define a cost function. Qubit drive and readout are then jointly optimized via gradient descent with finite differences and line search.

With this protocol, we locate the qubit frequency within an initial uncertainty of at least five Rabi frequencies and the optimal drive amplitude within a typical range of experimental variation, while simultaneously optimizing readout for maximum contrast between the qubit ground and excited state.

This removes the need for manual intervention, efficiently bridging the gap between rough initial guesses and high-fidelity optimization and providing a low-overhead approach to autonomous calibration of future quantum processors.