超节点与集群网络系统研究员
Core
Research next-generation hyper-node and cluster network systems for AI large-scale training, low-latency inference, and Agentic AI, focusing on network topology, protocols, and communication acceleration.
Role type
Senior IC research engineer (AI cluster networking)
Builds
High-performance network fabrics and communication engines for AI training and inference clusters
Domain
AI Infrastructure / High-performance Computing / Network Systems
Deliverable
production ML models
Required skills
AI large-scale training, low-latency inference, Agentic AI, network topology design, routing and load balancing, flow control, RoCE protocol stack, DPU/CCU/SSU communication engines, hardware-aware communication optimization, distributed training frameworks, deep learning frameworks, parallel computing principles
Preferred skills
Top-tier conference publications (NeurIPS, ICML, ISCA, MICRO, etc.), top research institution internships, ACM competition awards, NVIDIA GPU or Ascend NPU programming expertise
Technologies
PyTorch, vLLM, SGLang, VERL, NVIDIA GPU, Ascend NPU
Responsibilities
Design high-cost-performance network topologies and routing/flow control systems; Research and optimize protocol stacks (RoCE, custom bus protocols) for physical/link/network/transport layers; Innovate network IO and Fabric through hardware-software co-design; Optimize hardware-aware communication for large model training and inference; Build simulation platforms and validate prototype systems
Seniority
Senior, hands-on IC
