“The amount of data as well requires a massive telemetry apparatus, and designing likely models seems the province of people with PhDs.” Transfer learning take

“The amount of data as well requires a massive telemetry apparatus, and designing likely models seems the province of people with PhDs.” Transfer learning takes care of your first point. The second is taken care of by the plethora of open source libraries such as fast.ai that abstract away most of the decisions one needs to make when creating a deep learning model. Oh did I also mention Kaggle and Google Collab offer free GPU to train your models, in addition to the thousands of publicly available datasets?