LLM Training Architect: 1k GPU Distributed Systems
Posted 1
view
Summary
Onsite engineer in Gladstone, Queensland who orchestrates large-scale LLM pre-training runs across a ~1,000-GPU cluster, optimizing InfiniBand/RDMA networking and memory management and automating checkpointing and failure recovery. Core stack: C++, CUDA, and Python.
Hyphen Connect is seeking a highly skilled LLM Pre-training & Distributed Systems Engineer to orchestrate large-scale training runs and optimize distributed infrastructure. The role focuses on managing 1,000 GPUs and ensuring efficient, reliable training across advanced AI workloads at Gladstone, Queensland, AU. You will optimize InfiniBand/RDMA networking, memory management, and automate checkpointing and failure recovery during long-running jobs, drawing on strong C++, CUDA, and Python J-1880…