The real story here is that domain expertise can lead to superior outcomes. The hybrid curation pipeline still worked, but it was a low risk and low reward solu

The real story here is that domain expertise can lead to superior outcomes. The hybrid curation pipeline still worked, but it was a low risk and low reward solution that would eventually reach the goal. In general, there's a lot of unanswered questions regarding how to optimize in the risk reward ML space.