高性能算子开发工程师（自研芯片方向）（J97205）
Core
Develop and optimize operators for large language models on Baidu's self-developed chips to maximize compute performance and clear model dependencies.
Role type
Senior IC machine-learning engineer (self-developed chip operators)
Builds
Production ML operators and APIs for training and inference engines
Domain
AI hardware acceleration / Large Language Models
Deliverable
production ML models
Required skills
C/C++, Python, Linux, CUDA, OpenCL, AscendC, quantization techniques, model parallelism (TP/PP/DP), AI chip architecture (GPU/XPU/TPU/NPU), LLM architecture (Qwen/DeepSeek)
Preferred skills
Deep learning frameworks (PyTorch, TensorFlow, PaddlePaddle), AI Agent development
Responsibilities
Analyze performance bottlenecks of large models on GPU and self-developed chips; Develop operators for common large models based on Baidu's self-developed chips; Encapsulate and develop operator APIs for training and inference engines; Address operator dependencies for cutting-edge models; Explore using AI Agents to optimize operator performance
