GPU Kernel & Performance Engineer — CUDA/Triton Expert

CommonAI C.I.C.; Cambridge; On-site; GBP 120,000 - 170,000; Full time

Benefits offered by this job

Competitive salary

Pension

Professional development

Office near Cambridge station

Job summary

CommonAI CIC is seeking a GPU Kernel and Performance Engineer to accelerate research models on real hardware, primarily NVIDIA GPUs. You will join a compact team turning new model designs into efficient GPU code, measuring time and memory traffic on real hardware and expanding to other accelerators as needed.

You will profile, test, and optimise kernels, collaborating with training infrastructure to boost throughput and with researchers to validate fast designs.

Qualifications

• Proven experience writing and optimising GPU kernels in CUDA or Triton.
• Strong understanding of GPU architecture and memory hierarchies.
• Hands-on experience profiling and optimising performance with NVIDIA Nsight Compute or Nsight Systems.
• Strong Python and C/C++ skills.
• Rigorous testing of numerical code against reference implementations.

Responsibilities

• Write and optimise CUDA and Triton kernels for NVIDIA GPUs.
• Adapt and extend open-source kernels to new model designs.
• Profile training and inference to identify time and memory bottlenecks.
• Optimise for memory hierarchy and multi-GPU setups.
• Build tooling to measure data movement per token and validate cost models.

Skills

CUDA kernels

Triton kernels

GPU architecture

Nsight Compute

Nsight Systems

Python

C/C++

Testing numerical code

LLMs enthusiasm

Tools

NVIDIA Nsight Compute

NVIDIA Nsight Systems

CuTe

CUTLASS

Job description

CommonAI CIC is seeking a GPU Kernel and Performance Engineer to accelerate research models on real hardware, primarily NVIDIA GPUs. You will join a compact team turning new model designs into efficient GPU code, measuring time and memory traffic on real hardware and expanding to other accelerators as needed.

You will profile, test, and optimise kernels, collaborating with training infrastructure to boost throughput and with researchers to validate fast designs.