Senior Infrastructure Engineer in multi-GPU systems

About the position

The Senior Infrastructure Engineer will be responsible for architecting and deploying high-performance computing clusters with multi-GPU support specifically for AI/ML workloads. This role involves optimizing GPU resource scheduling, implementing infrastructure automation, and managing container orchestration systems to ensure efficient operation of GPU-accelerated workloads. The engineer will also focus on performance tuning, security compliance, and collaboration with MLOps teams to integrate GPU clusters into CI/CD pipelines.

Responsibilities

• Architect and deploy high-performance computing clusters with multi-GPU support for AI/ML workloads.
• Implement and optimize GPU resource scheduling, job queuing, and distributed training setups.
• Leverage NVIDIA CUDA to optimize performance for AI/ML models and workloads.
• Fine-tune GPU configurations for multi-GPU systems, ensuring maximum throughput and minimal latency.
• Build Infrastructure as Code (IaC) solutions using Terraform to automate the provisioning and management of on-premise infrastructure.
• Create scalable templates for consistent resource deployment.
• Deploy and manage container orchestration systems (e.g., Kubernetes, Docker Swarm) to run scalable GPU-accelerated workloads.
• Monitor and troubleshoot issues in distributed systems with tools like NVIDIA DCGM, Prometheus, or similar.
• Optimize AI/ML pipelines for distributed training across multi-GPU nodes.
• Develop strategies to efficiently utilize NVLink, NCCL, and other NVIDIA technologies.
• Set up robust monitoring and alerting systems to track GPU utilization, node health, and workload performance.
• Collaborate with MLOps teams to integrate GPU clusters into CI/CD pipelines.
• Implement security best practices for sensitive AI/ML workloads in an on-premise environment.
• Ensure compliance with organizational policies and industry standards.

Requirements

• 7+ years in infrastructure engineering, with at least 5 years of direct experience in GPU-accelerated systems and NVIDIA CUDA.
• Proven experience in deploying and managing multi-GPU systems for AI/ML workloads.
• Proficiency with NVIDIA CUDA for GPU programming and performance tuning.
• Hands-on experience with NVIDIA tools and libraries, including NVLink, NCCL, and cuDNN.
• Familiarity with MIG (Multi-Instance GPU) configurations and multi-GPU scaling techniques.
• Advanced knowledge of Terraform and scripting languages like Python or Bash for automation.
• Proficiency with container orchestration tools like Kubernetes or similar.
• Expertise in workload management systems and GPU monitoring tools (e.g., NVIDIA DCGM, Slurm).
• Experience in deploying and optimizing distributed training frameworks (e.g., TensorFlow MultiWorkerMirroredStrategy, PyTorch DDP).
• Strong understanding of networking, storage, and system architecture for high-performance compute environments.

Nice-to-haves

• Strong problem-solving abilities and critical thinking skills.
• Excellent communication skills for cross-functional collaboration.
• Leadership capabilities to guide junior engineers and manage projects.