Dr. Narayanan's research focuses on power-aware computer systems. Excessive power consumption is a key limiter to the design of modern computer systems ranging from large data centers to tiny and pervasive embedded internet of things. His research has addressed the power challenges in several intersecting domains such as the design of power-efficient computing systems, the design of application specific processors, the design of multicore architectures using emerging technologies like 3D stacking and nanotechnology, and the development of methodologies to understand interplay between performance, power and reliability.

Dr. Narayanan’s research has involved extensive interdisciplinary and multi-institutional collaborations. He works on developing intelligent vision systems that assist humans in civilian, defense and commercial applications. In the DARPA Neovision program, his team demonstrated brain-inspired vision systems that reduced power consumption of contemporary state-of-the-art by two orders of magnitude. Students trained in this project, currently lead commercial AI system design in Google, IBM, Intel and Meta. The commercial and societal impact of embedded vision systems for commercial (shopper insights), defense (autonomous weapons) and societal impact (assistive technologies) developed by Narayanan are cited in his election as Fellow of National Academy of Inventors. His work on an NSF expeditions-in-computing project resulted in simultaneous advances in computer vision, hardware, human-machine interface and neuroscience to enable a new generation of assistive vision technologies that enable independent shopping for persons with visual impairments. His team continues to work on developing assistive wearable vision systems supported by a NIH and an NSF project. His team is also working with entomologists in building sensor systems to track biodiversity in a NSF NRT project. 

Narayanan’s current work involves partnership with leading semiconductor companies (Intel, IBM. Global Foundries, Samsung and Micron) towards replacing the 50 year von Neumann computing paradigm that treats memory and processing distinct through novel processing-in-memory systems (JUMP 2.0 PRISM center). He is also collaborating with material scientists in the DoE 3DFeM center in developing Ferroelectric Memories. 
Efficient memory technologies are critical to reduce energy needs of emerging machine learning systems.