Senior GPU Kernel Engineer for Self-Improving AI Systems

Geometric; Greater London; Hybrid; GBP 90,000 - 140,000; Full time

Benefits offered by this job

Founding equity

Hybrid/remote flexible

Competitive salary

Job summary

Geometric is seeking exceptionally talented engineers and researchers to join us at the intersection of LLMs and evolutionary computing. You will write state-of-the-art GPU kernels and own complex production ML/AI systems end-to-end.

You will translate kernel-level gains into wall-clock improvements and build infrastructure for multi-day unsupervised runs, including design of the evolutionary search and sharing ideas through documentation and blogs.

Qualifications

• High-performance CUDA kernel development experience.
• Deep understanding of mixed precision and quantisation techniques.
• Experience with GPU tiling, memory access patterns and scheduling.
• Proven ability to trace perf cliffs via profilers.
• Familiarity with CuTe, Triton, Helion or PTX.
• Understanding GPU architecture across generations.
• Experience with production ML/training frameworks like Megatron-LM.
• Experience building performance-critical infra (compilers, profilers, auto-tuners).
• Interest in evolutionary methods and fitness landscapes.
• Knowledge of Neural Algorithmic Reasoning / Geometric DL.

Responsibilities

• Write SOTA GPU kernels.
• Own complex production ML/AI systems end-to-end.
• Understand how kernel-level gains translate to wall-clock improvements in production.
• Build the infrastructure enabling unsupervised multi-day iterations.
• Design the evolutionary search - fitness landscapes, operators, selection pressure.
• Communicate and share ideas through documentation, meet-ups and blogs.

Skills

CUDA kernels

GPU architecture knowledge

Mixed precision & quantisation

Profiling & optimisation

Transformers implementation

Production ML frameworks

Evolutionary methods intuition

Neural Algorithmic Reasoning

Geometric Deep Learning

GPU kernel performance engineering

Tools

CuTe

Triton

Helion

Job description

Geometric is seeking exceptionally talented engineers and researchers to join us at the intersection of LLMs and evolutionary computing. You will write state-of-the-art GPU kernels and own complex production ML/AI systems end-to-end.

You will translate kernel-level gains into wall-clock improvements and build infrastructure for multi-day unsupervised runs, including design of the evolutionary search and sharing ideas through documentation and blogs.