大模型系统性能/成本优化专家
Core
End-to-end performance and cost optimization for large model training and inference, focusing on low-precision training and end-cloud collaborative inference architectures.
Role type
Senior IC large model system performance and cost optimization engineer
Builds
End-to-end AI system solutions for typical business scenarios
Domain
AI systems, large language models, Ascend computing platform
Deliverable
production ML models
Required skills
Large model architecture (Transformer), low-precision training (mixed precision, quantization), operator optimization, distributed training frameworks, end-cloud collaborative architecture design, heterogeneous resource management, dynamic scheduling
Preferred skills
Ascend/GPU hardware optimization experience
Technologies
PyTorch, TensorFlow, Ascend platform
Responsibilities
Track frontier technologies for large model training and inference; build low-precision training technology systems; optimize large model inference performance on Ascend platform; design and implement cost-effective end-to-end AI system solutions
Seniority
Senior, hands-on IC
