I would say a PhD (or masters degree minimum) is required if you are developing new novel architectures, but not to iterate on existing architectures for a part

I would say a PhD (or masters degree minimum) is required if you are developing new novel architectures, but not to iterate on existing architectures for a particular application. Applied Data Science does not require developing as yet unseen architectures, being able to read cutting edge papers and applying the research is more than likely to enough.