Delay environment pod after agent pod

Due to the structure of my tensorflow code, I need to load the python libraries and model first before create the environment in the agent. Is it possible for our host to change the way environment pod is launched in evaluation scenarios?

Thanks!