GPU Systems Engineer for AI Inference & Performance

Nebius; United States; Remote; USD 180,000 - 260,000; Full time

Benefits offered by this job

Competitive compensation

Career growth

Flexible ownership & autonomy

Collaborative culture

Impactful AI projects

International environment

Job summary

Nebius is building Token Factory within Nebius Cloud, a world-scale GPU cloud for AI inference. The team develops and optimizes low-level kernels and runtime components to support fast, reliable deployment of foundation models across diverse hardware.

The role emphasizes performance tuning, profiling, and collaboration with ML and backend engineers to improve end-to-end execution on GPU platforms, with strong memory hierarchy and system-level expertise required.

Qualifications

• Strong proficiency in C++ and memory management for high-performance systems.
• Experience in GPU programming or systems-level software development.
• Hands-on profiling and debugging to optimize CPU/GPU performance.
• Knowledge of CPU/GPU architecture and memory hierarchy.

Responsibilities

• Develop and optimize low-level kernels and runtime components for AI inference.
• Improve performance of inference engines on GPU platforms.
• Profile and debug system-level and hardware-level performance issues.
• Integrate support for new hardware architectures (e.g., Hopper, Blackwell, Rubin).
• Collaborate with ML and backend teams to optimize end-to-end execution.

Skills

C++ programming

GPU programming

Kernel modules

Performance profiling

Tools

Perf tools

Nsight

VTune

ROCm profiler

Job description

Nebius is building Token Factory within Nebius Cloud, a world-scale GPU cloud for AI inference. The team develops and optimizes low-level kernels and runtime components to support fast, reliable deployment of foundation models across diverse hardware.

The role emphasizes performance tuning, profiling, and collaboration with ML and backend engineers to improve end-to-end execution on GPU platforms, with strong memory hierarchy and system-level expertise required.