The Power and Limits of Deep Learning (Hardware related Question)

I am a research assistant and I am very interested in the field of Deep Learning, but as I am a hardware guy, I want to know how to apply the concepts of deep learning in the Hardware?
In other words, I am doing a computer architecture where basically I am trying to use the emerging memory technology to enable the processing inside the memory as a promising solution for the memory-wall challenge. So my question is how to apply deep learning concepts in these systems?

Moreover, I want also to ask you, as a computer architect, how can I help to overcome the challenges of Deep Learning? … I mean, what can Hardware Researchers do to help in the Deep Learning field?

As per my knowledge , going forward deep learning layers need optimized and tuned HW due to huge number of learned parameters . Hence its important to understand neural nets and their architectures and understand how one can map it to hardware . Recent launch of Intel neural networks are examples of this .