Senior GPU/InfiniBand Systems Engineer for HPC Cloud

Nebius Group; Greater London; On-site; GBP 128,000 - 225,000; Full time

Benefits offered by this job

Competitive pay

Career growth

Flexibility

Collaborative culture

AI project impact

Global teams

Job summary

Nebius is seeking a Senior Software Systems Engineer to advance our hyperscaler platform. You will optimize GPU clusters, InfiniBand networks, and the KVM/QEMU stack, collaborating with hardware virtualization teams to ensure high performance and security in multi-GPU HPC environments.

You will analyze, troubleshoot, and improve infrastructure to support new hardware, automate fault detection, and drive performance improvements across the GPU/InfiniBand stack.

Qualifications

• Experience in system-level software development with performance focus.
• Strong Linux administration, troubleshooting and tuning skills.
• Solid understanding of server architecture, PCIe devices and HPC systems.

Responsibilities

• Tune performance of GPU clusters and InfiniBand networks in HPC environments.
• Troubleshoot root causes related to GPUs and InfiniBand networks and propose fixes.
• Integrate new hardware into infrastructure, supporting GPU hardware via software stacks like Kubernetes, QEMU, and KVM.
• Enhance automation for proactive monitoring and fault detection in GPU/InfiniBand systems.
• Configure and manage GPU devices and InfiniBand fabrics for reliable operation.

Skills

System-level software development

Performance optimization

C/C++

Go

Python

Linux

HPC

GPU programming

InfiniBand

PCIe

Education

Bachelor's degree in Computer Science

Tools

KVM/QEMU

Kubernetes

RDMA

RoCE

MPI

NCCL

PyTorch

TensorFlow

PCIe

Job description

Nebius is seeking a Senior Software Systems Engineer to advance our hyperscaler platform. You will optimize GPU clusters, InfiniBand networks, and the KVM/QEMU stack, collaborating with hardware virtualization teams to ensure high performance and security in multi-GPU HPC environments.

You will analyze, troubleshoot, and improve infrastructure to support new hardware, automate fault detection, and drive performance improvements across the GPU/InfiniBand stack.