Brendan Roberts
Papers
1
Total Citations
5
H-Index
1
About
Brendan Roberts is a leading figure in computer architecture, with a focus on efficient hardware design for machine learning and robotics. His most impactful work centers on developing specialized system-on-chips (SoCs) that bridge the gap between computation and memory. Roberts’s landmark paper, “NeCTAr and RASoC: Tale of Two Class SoCs for Language Model Interference and Robotics in Intel 16,” introduces the NeCTAr (Near-Cache Transformer Accelerator), a 16nm heterogeneous multicore RISC-V SoC that integrates both near-core and near-memory accelerators for sparse and dense machine learning kernels. This prototype chip, operating at 400MHz and 0.85V, achieves 109 GOPS for matrix-vector multiplications, demonstrating a significant leap in energy-efficient inference. With 5 citations in its first year, this work is already shaping the next generation of edge AI hardware. Roberts’s contributions are pivotal for enabling real-time, low-power language model and robotic applications, making him a key innovator in the field of domain-specific architectures.
Research Focus
Key Achievements
Top Papers
- 1