Sourjya Roy
Papers
1
Total Citations
41
H-Index
1
About
Sourjya Roy is a leading researcher at the intersection of computer architecture and deep learning, with a primary focus on compute-in-memory (CIM) technologies. His work addresses the critical challenge of energy-efficient AI inference by rethinking how neural network computations are mapped onto hardware. Roy’s most cited paper, “Compute-in-Memory Technologies and Architectures for Deep Learning Workloads” (2022), has garnered 41 citations, establishing him as a key voice in this rapidly evolving field. In this work, he systematically analyzes the virtuous cycle between algorithms, data, and computing capacity that has driven deep learning’s success, while identifying the memory wall as a fundamental bottleneck. By proposing novel CIM architectures that perform computations directly within memory arrays, Roy’s research offers a path to dramatically reduce data movement energy—a critical requirement for edge and embedded AI systems. His contributions are particularly notable for bridging the gap between device-level innovations and system-level design, making his work essential reading for researchers and engineers working on next-generation accelerators for computer vision, speech recognition, and robotics.
Research Focus
Key Achievements
Top Papers
- 1Compute-in-Memory Technologies and Architectures for Deep Learning Workloads41 citations · 2022