大模型训练研发工程师（J95973）
Core
Develop training optimization for Baidu's ERNIE Bot large language model and distributed training architecture for the PaddlePaddle deep learning framework.
Role type
Senior IC machine-learning engineer (large model training & distributed systems)
Builds
Large-scale distributed training infrastructure, high-performance computing libraries, and communication libraries for deep learning frameworks.
Domain
Artificial Intelligence / Deep Learning / High-Performance Computing
Deliverable
production ML models
Required skills
CUDA programming, distributed training architecture, high-performance computing, Linux/Unix development, multi-threaded programming, network programming, algorithm-engineering co-optimization
Preferred skills
PaddlePaddle framework experience, DeepSpeed, Megatron, cross-modal model optimization
Technologies
CUDA, Linux, PaddlePaddle, DeepSpeed, Megatron
Responsibilities
Design and develop high-performance computing platforms and communication libraries; explore algorithm-engineering co-optimization for large language and cross-modal models; optimize deep learning frameworks to reduce development costs; perform high-quality development, self-testing, and documentation for training projects.
Seniority
Senior, hands-on IC