I am an EECS student at the University of Michigan with research interests in computer architectures for robotics and computer vision applications, in/near-sensor computing, and computing with novel devices.
Temporal In-Sensor Computing for Robot Localization
May 2023 - Present
University of Michigan - Ann Arbor
Advised by Prof. George Tzimpragos
- The high energy cost of data movement from the sensor to the processor prompted a shift towards moving computation closer to the sensor itself. This work focuses on using race logic, which encodes information as timing delays, to enable more energy-efficient computation for in-sensor processing. We find that temporal logic is a natural fit for feature detection and tracking in robot localization, and the use of race logic allows us to simplify the sensor interface of CMOS image sensors.
Superconducting Neural Network Accelerator
Nov 2022 - Present
University of Michigan - Ann Arbor
Advised by Prof. George Tzimpragos
- Superconductor electronics have been demonstrated to be a promising alternative to CMOS for building neural network accelerators due to their ultra-fast speed and high energy efficiency. However, the characteristics of SFQ devices presents several architectural limitations, including the need for gate-level clocking and the lack of fast and reliable random access memory. This work aims to leverage newly proposed SFQ logic schemes and technologies, namely xSFQ logic and delay line memories, to design more efficient superconducting neural network accelerators.