The Role:
We are looking for a principal compute architect to lead a small group of outstanding HW and SW architects to innovate and co-develop features and changes to our parallel compute and deep learning hardware system that will shape efficiency.
Key Responsibilities:
· Analyze, Architect, Model and analyze resulting hardware and software codesigned models for new features to improve asynchronous multi-task work scheduling, GPGPU based deep learning accelerators.
· Identify emerging techniques and NN models / algorithms to guide compute system & ISA & AI compiler designs
· Analyze operations to identify actual performance bottlenecks in hardware and/or software on both silicon or pre-silicon GPGPUs
· Work with software, hardware, architecture, and design implementation teams to identify issues and improve performance
任职要求
Preferred Experience:
· GPU driver understanding/experience with GPGPU compute programming
· Knowledge in neural network models especially LLMs and AI compiler and Operators Lib
· Multi-Core, Multi-Thread Architectures enabling performant irregular workloads
· Strong programming & hardware modeling skills in C/C++ (GPGPU Sim or Multi-sim a plus)
· Scripting language (Python, Perl) and UNIX/Linux experience, RTL a plus
· OpenCL/CUDA/ROCm development experience desired
Academic Credentials:
· Masters, PhD & 10 years proven experience in relevant field (CS, EE, CE, Math, Physics)
