AI infra研发工程师(北京/深圳)
Core
Develop high-performance inference and training frameworks based on self-developed chips to solve full-link issues in chip deployment.
Role type
Senior IC AI infrastructure engineer (chip software stack)
Builds
Self-developed chip software ecosystem and optimized inference/training frameworks
Domain
AI infrastructure, custom silicon, high-performance computing
Deliverable
production ML models
Required skills
Linux development, Python, C++, GPU/SIMD programming, LLM/AIGC model architecture, profiling tools (nvpro, nsys), training/inference frameworks (PyTorch, Megatron, DeepSpeed, VLLM, Sglang), operator-level performance analysis
Preferred skills
System design, full-stack optimization of inference networks
Technologies
PyTorch, Megatron, DeepSpeed, VLLM, Sglang, nvpro, nsys
Responsibilities
Develop high-performance inference and training frameworks based on self-developed chips; Iterate framework performance and usability based on hardware and model characteristics; Collaborate with business teams to build the self-developed chip software ecosystem
Seniority
Senior, hands-on IC
