Tushar Krishna
Papers
4
Total Citations
596
H-Index
3
About
Tushar Krishna is a leading researcher at the intersection of computer architecture and deep learning, whose work is fundamentally reshaping how neural networks are accelerated in hardware. His primary research areas include hardware accelerators for deep learning, on-chip networks, and efficient computing systems. Krishna’s most impactful contribution is the SIGMA accelerator, which introduced a flexible, sparse-aware architecture for irregular GEMM operations—a critical bottleneck in DNN training. With 462 citations, SIGMA has become a foundational reference for designing accelerators that handle sparsity and irregular dataflow, enabling more efficient training of large-scale models. He has also made significant strides in characterizing DNN deployment on commercial edge devices (109 citations), providing vital insights into the performance and energy trade-offs of inference at the edge. Additionally, his work on GeneSys explores the hardware support for continuous learning through neural network evolution, pushing the boundaries of adaptive AI systems. Krishna’s research not only advances the theoretical understanding of hardware-software co-design but also delivers practical solutions that directly impact the efficiency of modern AI workloads, making him a pivotal figure in the field.
Research Focus
Key Achievements
Top Papers
- 1
- 2Characterizing the Deployment of Deep Neural Networks on Commercial Edge Devices109 citations · 2019
- 3
- 4