AI 通信研发工程师（J100998）
Core
Optimize communication system performance for AI training and inference scenarios, focusing on large-scale distributed systems and MoE architectures.
Role type
Senior IC AI Infrastructure Engineer (Communication Systems)
Builds
High-performance communication stacks, runtime systems, and optimization solutions for AI clusters.
Domain
AI Infrastructure / Distributed Systems / High-Performance Computing
Deliverable
production ML models
Required skills
C/C++ development, Linux system programming, GPU communication principles, MoE architecture understanding, distributed training frameworks, cluster stability analysis
Preferred skills
NCCL/DeepEP/Mooncake expertise, heterogeneous network support, fault recovery mechanisms, performance benchmarking
Technologies
NCCL, DeepEP, Mooncake, MPI, Gloo, PyTorch Distributed, Megatron-LM, DeepSpeed, vLLM, SGLang
Responsibilities
Analyze and optimize communication performance in training/inference; design All-to-All and Token Dispatch/Combine schemes; troubleshoot system-level issues like hangs and slow nodes; develop self-built communication frameworks; collaborate across model, framework, and hardware teams; establish performance baselines and reporting.
Seniority
Senior, hands-on IC
