AI Training GPU Performance Engineer

present of AI; Cambridge; Remote; GBP 55,000 - 90,000; Full time

Job summary

CuspAI is seeking a performance engineer to optimize large AI training on GPU clusters. You will profile workloads, tune CUDA/Triton kernels, and optimize inter-node communication while building dashboards to show hardware efficiency.

The role suits someone with hands-on HPC experience, comfortable working in a young company, and eager to directly influence AI research throughput and cost efficiency.

Qualifications

• Hands-on performance or HPC work experience required.
• Strong knowledge of CUDA or Triton.
• Ability to work autonomously in a small, fast-growing company.

Responsibilities

• Profile large AI training runs on GPU clusters to identify wasteful compute.
• Tune GPU kernels and inter-machine communication for better efficiency.
• Build dashboards that visualize hardware utilization and bottlenecks.

Skills

Profiling

HPC work

CUDA

Triton

Performance tuning

Education

B.Sc. in CS/EE or equivalent

Tools

CUDA toolkit

NVIDIA profiling tools

Job description

CuspAI is seeking a performance engineer to optimize large AI training on GPU clusters. You will profile workloads, tune CUDA/Triton kernels, and optimize inter-node communication while building dashboards to show hardware efficiency.

The role suits someone with hands-on HPC experience, comfortable working in a young company, and eager to directly influence AI research throughput and cost efficiency.