AI Infrastructure Engineer: GPU & Performance Optimisation

twentyAI; Greater London; On-site; GBP 90,000 - 140,000; Full time

Job summary

twentyAI is seeking an AI Infrastructure Engineer to build autonomous systems that optimise modern AI workloads. You will work across GPU programming, ML infrastructure, and automated optimisation to tackle challenging performance problems with real production impact.

The role involves designing, deploying and tuning production AI infrastructure, measuring end-to-end efficiency, and contributing to open-source projects. Lead candidates may mentor a small team.

Qualifications

• Experience building performance-critical software for GPU-accelerated environments.
• Deep understanding of GPU performance optimisation and memory usage.
• Experience with mixed-precision and quantised AI workloads (INT4, INT8, FP8).
• Understanding of distributed AI systems and performance techniques.
• Ability to diagnose bottlenecks using modern profiling tools.

Responsibilities

• Develop, optimise and deploy performance-critical software for modern AI workloads, including low-level GPU acceleration.
• Own production ML infrastructure and AI systems from design to deployment.
• Measure how performance improvements affect end-to-end efficiency.
• Build infrastructure for large-scale optimisation experiments with tracking and benchmarking.
• Design automated optimisation and search strategies to improve system performance.
• Investigate bottlenecks and translate findings into engineering improvements.
• Share knowledge via documentation, talks and open-source contributions.
• For lead candidates, hire and mentor a small team.

Skills

GPU programming

Performance optimization

Profiling tools

Distributed AI systems

Transformer architectures

Open-source contributions

Research publications

Tools

CUDA

Triton

CuTe

Helion

Megatron-LM

vLLM

Job description

twentyAI is seeking an AI Infrastructure Engineer to build autonomous systems that optimise modern AI workloads. You will work across GPU programming, ML infrastructure, and automated optimisation to tackle challenging performance problems with real production impact.

The role involves designing, deploying and tuning production AI infrastructure, measuring end-to-end efficiency, and contributing to open-source projects. Lead candidates may mentor a small team.