AI推理加速方向工程师（J106319）
Core
Optimizing inference engines for large-scale MoE architectures (TB-level) and multi-modal scenarios (text, image, video, audio) to achieve extreme performance, low latency, and cost efficiency.
Role type
Senior IC AI inference acceleration engineer (LLM & multi-modal)
Builds
High-performance inference engines, multi-chip acceleration solutions, and optimized operator libraries for GPU and heterogeneous chips.
Domain
Artificial Intelligence / Large Language Models / Heterogeneous Computing
Deliverable
production ML models
Required skills
LLM inference engine optimization, heterogeneous computing programming, CUDA development, high-performance operator development, model deployment and optimization, service scheduling, memory management, model quantization and sparsification, multi-modal architecture design, toolchain engineering (Harness/Loop/Graph).
Preferred skills
Experience with domestic/heterogeneous chips (Kunlun, Ascend, AMD), community contributions to vLLM/SGLang/FlashInfer, algorithm research for SOTA models.
