calculating hybrid beamforming parameters for communications with a second device based at least in part on an estimation of channel information associated with at least a portion of a communication link at the second device via a machine learning model, wherein inputs to the machine learning model comprise measured parameters based at least in part on measurements performed by the second device, by one or more other devices, or both;

transmitting, to the second device, a message indicating the hybrid beamforming parameters, wherein the hybrid beamforming parameters comprise an input for a beam refinement process at the second device; and

communicating with the second device using one or more beams based at least in part on the hybrid beamforming parameters and a result of the beam refinement process.