Tianmu Li and Jiyue Yang, PhD students in the UCLA Electrical and Computer Engineering department, were awarded the 2021 Qualcomm Innovation Fellowship, a competitive program that celebrates PhD students’ innovative research and discoveries.
While the program spans across the globe, Li and Yang were selected as one of 16 teams out of over 100 submissions to be granted the fellowship in North America. Their winning proposal, “A Stochastic Compute-In-Memory Neural Network Accelerator with Variable Precision Tuning,” was the only one chosen from UCLA.
Yang summarizes the proposal as a way to focus on “building a very low-power neural network accelerator for IoT [Internet of Things] edge devices.” A significant number of IoT edge devices use small batteries as a source of energy, which can prove troublesome for “large deep-learning algorithms” that require more energy than these batteries can provide. The proposal seeks to change this by addressing how deep-learning algorithms could be computed more efficiently. The implications for the design would “give edge devices more compute power and provide users faster responses and make their data more private.”
Such a proposal requires an enormous amount of teamwork and communication between Li and Yang, who explained how they approached working on the proposal together. “I [Yang] am focusing on designing and building high-performance circuit hardware. Tianmu is focusing on training and improving deep-learning algorithms in software. We have to constantly discuss with each other to make sure that the hardware and algorithms are not only compatible but also bring out the best potential performance.”
The one-year fellowship offers an abundance of benefits, including exclusive mentorship from Qualcomm’s engineers, as well as $100k in fellowship funding. Li and Yang have already planned the next steps of their proposal, which includes drafting and producing silicon chip hardware for their project.
This consistent effort and planning have provided Li, who is Professor Puneet Gupta’s PhD advisee, and Yang, who is Professor Sudhakar Pamarti’s PhD advisee, with an unparalleled chance to continue developing their skills as engineers as they learn from some of the best in the world. Yang noted, “This fellowship is not only a recognition of our work from a leading company in the semiconductor industry, but also a good opportunity for us to exchange ideas with Qualcomm’s senior engineers.”