AI for Science性能优化工程师-Seed
Core
Optimizing training and inference performance for AI4S models (e.g., MLFF, Cofolding) to improve throughput, resource utilization, and computational efficiency.
Role type
Senior IC AI system performance engineer
Builds
High-performance AI4S model training and inference pipelines
Domain
Artificial Intelligence for Science (AI4S) / High-performance Computing
Deliverable
production ML models
Required skills
Python, C++, PyTorch, CUDA programming, GPU architecture knowledge, performance profiling tools, Triton, Torch.Compile, distributed training frameworks
Preferred skills
CUDA Kernel development, compilation optimization, TorchInductor, TensorRT, XLA, TVM, large model optimization, scientific computing workloads
Responsibilities
Analyze and resolve performance bottlenecks in model training and inference; implement operator fusion and compilation optimizations; establish performance analysis and benchmarking systems; collaborate with algorithm and platform teams to drive infrastructure evolution.
