训练高性能计算工程师 - Seed Model
Core
Design and implement key high-performance operators for large-scale training systems, optimizing performance across AI accelerators and distributed parallelism for foundational models.
Role type
Senior IC high-performance computing engineer (seed model training)
Builds
High-performance training systems for foundational models (MLLM, GenMedia, AI for Science)
Domain
AI infrastructure / Large-scale model training
Deliverable
production ML models
Required skills
C/C++, Python, PyTorch, CUDA programming, Triton, TVM, Torch.compile, NCCL, NVSHMEM, RDMA, NVLINK, large model architectures
Preferred skills
CUTLASS, Flash Attention optimization, Tilelang, compute/communication overlap operators
Responsibilities
Design and implement critical high-performance operators; Optimize operator performance for specific AI accelerator architectures; Implement optimizations for distributed computing and communication; Collaborate on algorithm-engineering design for optimal system performance and model quality.