Role Overview

Develop key components and core technologies that enable various (based-on-SOTA) AI models to run efficiently on NPUs.

Research NPU hardware architecture optimized for low power consumption and high performance.

Design data workloads that maximize energy efficiency of conventional NPUs.

Continuously evaluate and ensure alignment between user requirements and NPU specifications.

Conduct ongoing research and development on NPU hardware bottlenecks and breaking points.

Responsibilities

• Explore, discuss, and determine NPU architectures that meet required specifications.
• Collaborate with internal teams or customers to develop profiling tools tailored to specific needs.
• Predict hardware-friendly performance quickly and accurately for neural networks using C++ or SystemC.
• Work closely with deep learning, RTL, and compiler teams to design efficient engineering methodologies and N...