AscendC accelerator development anchors Huawei Canada’s compute operator work across FlashAttention, FP8 GEMM, fused normalization, and agile MoE routing.
As a Senior System Engineer, you will implement high-performance kernels for foundation-model workloads on Ascend NPUs. You will explore Triton and TileLang while developing operators with AscendC. The position also connects kernel performance with low-precision alignment, hardware architecture, and open-source acceleration.
Key Responsibilities: Implement AscendC kernels for core foundation-model operators
Explore Triton-Ascend and TileLang accelerator backends
Optimize FP8, MXFP8, W4A8, and W8A8 computations
Analyze Cube, Vector, memory, and HCCL performance behavior
Support open-source Ascend and AI acceleration projects
Requirements: Bachelor’s, Master’s, or PhD in a technical discipline
Three-plus years of relevant HPC or accelerator experience
AscendC, CANN, or transferable CUDA C/C++ expertise
Interest or experience in Triton or TileLang
Knowledge of low-precision numerical alignment and profiling
Develop AscendC kernels and hardware-aware optimizations for foundation models while advancing numerical precision and accelerator ecosystem capabilities.
#J-18808-Ljbffr
📌 Senior AscendC AI Accelerator Engineer (British Columbia)
🏢 Huawei Canada
📍 British Columbia
Reply to this offer
Impress this employer describing Your skills and abilities, fill out the form below and leave Your personal touch in the presentation letter.