Use Deep Learning project and remove workaround when fixed

[markstur]

There is a documented workaround for an auth issue w/ the Deep Learning project.
That should be removed once the problem is resolved.
Also I think we want this to run under the Deep Learning project but it is currently documented as a Data Science project. Does it matter? -- I think after that auth problem is fixed it'd make more sense to use Deep Learning (or is there a more minimal one that is preferred).