微信支付-大模型 AI Infra 研发工程师
Core
Building and optimizing the WeChat Pay large model inference and training platform, focusing on hardware-software co-optimization for heterogeneous GPU clusters.
Role type
Senior IC large model AI infrastructure engineer
Builds
Large-scale inference and training platforms for WeChat Pay's large models
Domain
AI Infrastructure / Large Language Models / Heterogeneous Computing
Deliverable
production ML models
Required skills
vLLM, SGLang, DeepSpeed, Megatron-LM, distributed training frameworks, GPU architecture, operator optimization, graph optimization, NCCL optimization, CUDA development, heterogeneous GPU adaptation
Preferred skills
Domestic GPU adaptation, AI compiler development, large-scale GPU cluster management, open-source project maintenance, ACM awards
Technologies
vLLM, SGLang, DeepSpeed, Megatron-LM, verl, LLaMA-Factory, CUDA, NCCL, DeepSeek, Hunyuan, Qwen, GLM, Kimi
Responsibilities
Build and optimize the large model inference platform for performance and throughput; Optimize large model training performance for SFT and reinforcement learning tasks; Adapt and optimize for domestic GPUs and heterogeneous computing resources
Seniority
Senior, hands-on IC