Passing data between assets when some use pipes

[eteran]

I am interested in having some assets be defined in separate processes, perhaps even implemented in other languages. My understand is that to do this, I have to use pipes as shown here: https://docs.dagster.io/guides/build/external-pipelines/

However, it isn't clear how (if at all) this works when we want to send/receive data from other assets.

For example:

given this simple pair of assets, where the first retrieves some data, and the second transforms it:

@dg.asset
def raw_data() -> str:
    # Code to fetch raw data from an external source
    return "Raw data"


@dg.asset(deps=[raw_data])
def transformed_data(raw_data) -> str:
    # Transform it
    return raw_data.upper()


defs = dg.Definitions(
    assets=[raw_data, transformed_data],
)

Suppose I want to change transformed_data to do the work in an external process. I'd switch it to use pipes_subprocess_client, right? But how would I pass the raw_data to that piped process and then returns it's stdout?

I suppose that I can literally just use the python subprocess module or similar and just do something like sending the raw data as a JSON object to stdin and look for a JSON object on stdout. But is there a "dagster way" to do things using pipes_subprocess_client while maintaining the ability to send/receive the data with simplicity?