Vivado Partial Reconfiguration (Dynamic Function Exchange)

I'm looking for an expert who can help me partially reconfigure an ML accelerator in Vivado. Although I'm not specific about the changes required, performance increase, memory efficiency improvements or/and supporting new hardware features may be necessary. The program is run on Pynq for image classification using a DNN. I am not settled on the classification task yet, so flexibility is crucial. Applicants should provide information about their relevant experience. Proficiency with ML accelerators, Vivado, Pynq and DNN is essential.