Research MLE (Training Optimization）大模型训练优化工程师
Core
Design, implement, and optimize large-scale distributed training systems for multimodal and foundation models to improve GPU utilization, communication overhead, and memory efficiency.
Role type
Senior IC machine learning engineer (training optimization)
Builds
Distributed training infrastructure for large-scale multimodal and foundation models
Domain
Generative AI, Large Language Models, Multimodal AI
Deliverable
production ML models
Required skills
Distributed training systems, GPU optimization, CUDA/Triton kernel development, PyTorch, Megatron-LM, NeMo, FSDP/ZeRO, gradient checkpointing, low-precision data types, Python, C++ or Rust
Preferred skills
Experience with diffusion models, system programming languages
Technologies
Megatron-LM, NVIDIA NeMo, FSDP, Triton, CUDA, PyTorch, DeepSpeed
Responsibilities
Design and optimize large-scale ML training systems, improve performance across compute/memory/communication layers, partner with research teams, debug and profile training workflows, write custom GPU kernels
Seniority
Senior, hands-on IC