The API exposes the chunk/delay parameters as options, so it's definitely possible.

If you mean having the library code "learn" the optimal chunk/delay for a given run, that's possible too, but I'd have to wonder about alternative solutions at that point.