Lead GPU Performance Systems Engineer

Nebius Group; United States; Remote; USD 170,000 - 300,000; Full time

Benefits offered by this job

Health insurance

401(k) plan

Parental leave

Remote work stipend

Disability & life insurance

Job summary

Nebius, a Nasdaq-listed Amsterdam-based cloud AI platform, is seeking a Lead Software Systems Engineer - GPU Performance to optimize large-scale GPU clusters across hardware and software layers.

You will work across InfiniBand/RoCE, KVM/QEMU, MPI/NCCL, and system software, collaborating with infra, software, and hardware partners to push performance, reliability, and scalability for training and inference workloads.

Qualifications

• 5+ years of professional experience in system-level software development with a focus on performance optimization.
• 3+ years hands-on experience with Linux systems (administration, troubleshooting, and performance tuning).
• Deep understanding of server architecture, PCIe devices, NICs, Linux OS/Kernel, and HPC systems.
• Strong proficiency in performance-oriented programming languages (C/C++, Go, Python).

Responsibilities

• Identify performance bottlenecks across multiple layers and drive improvements for large-scale GPU clusters.
• Troubleshoot performance issues under real workloads (training and inference).
• Evaluate and integrate new hardware, system configurations and tuning approaches through the software stack.
• Support complex performance-related escalations from internal teams and customers.
• Collaborate with infrastructure, software engineering and hardware vendor teams (e.g., NVIDIA, Mellanox, Intel).
• Contribute to hardware and cluster qualification to ensure performance expectations are met.

Skills

System-level software development

Linux systems

Performance optimization

C/C++

Tools

MPI

NCCL

KVM/QEMU

Job description

Nebius, a Nasdaq-listed Amsterdam-based cloud AI platform, is seeking a Lead Software Systems Engineer - GPU Performance to optimize large-scale GPU clusters across hardware and software layers.

You will work across InfiniBand/RoCE, KVM/QEMU, MPI/NCCL, and system software, collaborating with infra, software, and hardware partners to push performance, reliability, and scalability for training and inference workloads.