Bryan Ngo
Papers
1
Total Citations
5
H-Index
1
About
Bryan Ngo is a computer architect whose work bridges the gap between hardware acceleration and real-world AI deployment, with a focus on energy-efficient, domain-specific computing. His key research areas include near-memory processing, transformer accelerator design, and heterogeneous system-on-chip (SoC) architectures for machine learning and robotics. Ngo’s most notable contribution is the NeCTAr (Near-Cache Transformer Accelerator), a 16nm heterogeneous multicore RISC-V SoC introduced in his 2024 paper, which has already garnered 5 citations. This prototype chip, operating at 400MHz and 0.85V, achieves 109 GOPS for matrix-vector multiplications by integrating both near-core and near-memory accelerators to efficiently handle sparse and dense machine learning kernels. His work, co-developed with the RASoC team, demonstrates a pragmatic approach to overcoming memory bottlenecks in language model inference and robotic control. By pushing the boundaries of how accelerators interact with memory hierarchies, Ngo is shaping the next generation of low-power, high-performance edge AI hardware—a critical step toward making advanced AI accessible in resource-constrained environments.
Research Focus
Key Achievements
Top Papers
- 1