CUDA/Triton Kernel Engineer for Fast AI Inference

CommonAI Holdings Ltd; Cambridge; Hybrid; GBP 90,000 - 120,000; Full time

Benefits offered by this job

Competitive salary

Pension

Professional development

Networking opportunities

Vibrant Cambridge office

Job summary

CommonAI CIC in the UK is seeking a GPU Kernel and Performance Engineer to accelerate research models on NVIDIA GPUs. You will turn model designs into efficient CUDA/Triton code and measure memory traffic on real hardware, with a focus on performance at scale.

You will collaborate with a small research team, optimize for advanced memory hierarchies, and extend open-source kernels while exploring new AI accelerators as the project evolves.

Qualifications

• Proven experience writing and optimising GPU kernels in CUDA or Triton.
• Strong understanding of GPU memory hierarchies and architecture.
• Hands-on experience profiling with Nsight Compute or Nsight Systems.
• Strong Python and C/C++.
• Rigorous numerical testing and validation.
• Open to learning new hardware architectures and programming models.

Responsibilities

• Write and optimise CUDA and Triton kernels on NVIDIA GPUs for sparse and structured model components.
• Adapt and extend open-source kernels, such as FlashAttention-style and block-sparse attention, to new model designs.
• Profile training and inference to find where time and memory traffic go, using tools such as NVIDIA Nsight Compute and Nsight Systems.
• Optimise for the GH200's memory hierarchy on Isambard-AI and for our own GPU nodes.
• Build tools that measure data movement per generated token, and check analytic cost models against hardware measurements.
• Test every kernel against reference implementations before it runs at scale.
• Work with our training infrastructure engineer to raise training throughput, and with researchers to make new designs fast enough to test.
• Evaluate how our models map onto other AI accelerators, such as TPUs and dataflow, wafer-scale and compute-in-memory architectures, and port key components to them where it is useful.
• Contribute improvements back to open-source projects where possible.

Skills

CUDA kernels

Triton kernels

GPU architecture

Profiling tools

Python

C/C++

Numerical testing

Learning new hardware

LLMs/AI tooling

Tools

NVIDIA Nsight

CUTLASS

CuTe

NCCL

NVLink

Job description

CommonAI CIC in the UK is seeking a GPU Kernel and Performance Engineer to accelerate research models on NVIDIA GPUs. You will turn model designs into efficient CUDA/Triton code and measure memory traffic on real hardware, with a focus on performance at scale.

You will collaborate with a small research team, optimize for advanced memory hierarchies, and extend open-source kernels while exploring new AI accelerators as the project evolves.