Seems to be a no-brainer to me. The iteration speed in data science is so much faster if one person controls the entire workflow, at least for the alpha and beta release. Further system optimization can be done by more specialized workforce.