Researcher - AI Computing System
Core
Design and develop optimization solutions for AI training and inference systems on the Ascend platform, focusing on low-precision training, multimodal optimization, and reinforcement learning to address system bottlenecks.
Role type
Researcher - AI Computing System
Builds
Stable, efficient AI training clusters and software frameworks/operator libraries for NPU platforms
Domain
AI computing systems, large model training/inference, NPU hardware architecture
Deliverable
production ML models
Required skills
Low-precision training, multimodal reinforcement learning, system-level architectural optimization, NPU programming, dynamic cluster configuration, FP8 optimization
Preferred skills
Experience with large model structures (Deepseek, Llama), AI training framework development, algorithm hardware experience, research in new architectures
Technologies
Ascend platform, NPU, Python, C, C++, FP8, RL, LLM, MoE
Responsibilities
Design optimization solutions for AI training/inference systems; Build stable AI training clusters; Develop software frameworks and acceleration libraries for NPU platforms; Drive innovation in large-model training with low-precision and parallel strategies; Track latest research trends in AI computing cluster architecture
Seniority
Mid-Senior, hands-on IC