There have to be automated ways of scaling intern hiring that do not involve this ProctorU-based approach.

If you have thousands of data points, what makes a successful candidate, and possibly, build a model around that?