About this role
Research MLE (Training Optimization）大模型训练优化工程师 at Canva. Location: Beijing, Beijing, China. Role: designing systems, optimizing training, writing kernels Requirements: Experience with large-model distributed training, proficiency in Python, PyTorch/JAX, CUDA/Triton, familiarity with Megatron-LM/NeMo/DeepSpeed, strong systems and optimization skills, and full professional English. Category: Software Development Seniority: Mid Level Tools: Megatron-LM, NVIDIA NeMo, FSDP, Triton, Python, C++, Rust, PyTorch, JAX, DeepSpeed, CUDA Commitment: Full Time Workplace: Hybrid Languages: English