GPU Cluster Engineer (Mensch)

The role

You will be the central contact for NEURA's GPU cluster infrastructure, an AWS HyperPod setup for foundation model training and fine-tuning. You will develop operating frameworks, build self-service tools for ML teams, and work with AWS to shape the platform.

What you will do

• Set up, configure and continuously develop HyperPod clusters including HyperPod/Slurm and HyperPod/EKS orchestration models.
• Design and implement strategies for cluster stability including node failure detection and automatic job recovery.
• Build a workload priority framework to share cluster capacity between teams.
• Optimize end-to-end GPU utilization by resolving bottlenecks in compute, GPU memory, EFA network and storage throughput.
• Collaborate with AWS HyperPod product and engineering teams.
• Provide self-service tools for ML researchers and engineers to start, monitor and manage training jobs.
• Create onboarding documentation, training materials and internal workshops.
• Manage costs and capacity strategy including Spot management, Reserved Instance planning and Savings Plans.
• Apply Infrastructure as Code for all cluster configurations and changes.

Required qualifications

• 5+ years experience in infrastructure or systems engineering, ideally with focus on GPU clusters or HPC environments.
• Deep practical experience with AWS HyperPod and AWS instances.
• Solid understanding of Slurm and Kubernetes as orchestration layers.
• Practical knowledge of distributed training.
• Experience in developing self-service tools and technical documentation.
• Strong understanding of cloud cost management at scale.
• Very good English skills.

Nice to have

• Direct experience with HyperPod is a strong advantage.
• German is a plus.

About the company

neura-robotics-gmbh