Performance issue. how to reduce the data conversion overhead between python and matlab?

[KunHwang]

You know since 2021a the comm. performance between python and matlab is significantly improved.
but not yet satisfied.
Still, there is the payback of data changing in function call from python. For the frequent call between python and matlab, especially Reinforcement Learning problem from Maltab Environment to Python Deep Neural Network Agent, it is a critical performance issue. Is there any chance to increase the data change performance ? or is it the limit? any plan to increase the comm. performance? in next version matlab ? tell me more. any chance to improve the performance in current use ? (Im using r2024a)