GPU Engineer

Skillsearch; United Kingdom; Remote; GBP 50,000 - 70,000; Full time

Job summary

A global technology consultancy in the UK is seeking a skilled GPU Engineer to optimise cloud-based performance solutions. Responsibilities include benchmarking GPU workloads, developing automated templates, and integrating GPU resources into CI/CD pipelines. The ideal candidate has at least 2 years of relevant experience and strong expertise in GPU technologies. This role offers an opportunity to contribute to high-impact AI projects in a multicultural environment.

Qualifications

• Minimum of 2 years’ experience in GPU engineering or cloud-based GPU workload optimisation.
• Demonstrable expertise in GPU-accelerated development and profiling tools.
• Proficiency in scripting languages for automation.

Responsibilities

• Conduct benchmarking, profiling, and tuning of GPU workloads.
• Develop automated configuration templates and workflows.
• Integrate GPU provisioning into CI/CD pipelines.

Skills

GPU engineering expertise

CUDA

OpenCL

PyTorch

TensorFlow

Terraform

Kubernetes

Python

Bash

Problem-solving skills

Tools

Nsight

Prometheus

Grafana

NVIDIA DCGM

Job description

Overview

Are you a skilled GPU Engineer with expertise in optimising GPU-accelerated workloads and cloud-based performance solutions?

This is an opportunity to contribute to a high-impact project, supporting advanced AI and data-driven use cases by designing, profiling, and streamlining GPU workloads in an international organisation.

Responsibilities

• Conduct benchmarking, profiling, and tuning of GPU workloads to provide recommendations on optimal GPU sharing techniques such as MIG, vGPU, MPS, and time-sharing
• Optimise existing and new applications by leveraging GPU parallelisation, identifying bottlenecks, and implementing framework-level improvements
• Analyse deployment methods for GPU-accelerated serving frameworks, with reference implementations and best-practice recommendations (e.g. NVIDIA Triton Inference Server, TensorRT, ONNX Runtime)
• Develop automated configuration templates and workflows for GPU resources using Infrastructure as Code (IaC) tools such as Terraform and Helm
• Implement active GPU monitoring with dashboards and alerts covering utilisation, memory bandwidth, temperature, and power metrics
• Integrate GPU resource provisioning and configuration into CI/CD pipelines to support seamless deployment and rollback
• Support self-service deployment of Large Language Models (LLMs) on GPU resources for application owners
• Document all configurations, benchmarking results, and deployment procedures for transparency and reproducibility
• Deliver knowledge transfer and training sessions for ICT staff on GPU workload optimisation, management, and troubleshooting

What we're looking for

• Minimum of 2 years’ hands-on experience in GPU engineering or cloud-based GPU workload optimisation within enterprise or large-scale environments
• Demonstrable expertise in GPU-accelerated development using CUDA, OpenCL, PyTorch, TensorFlow, TensorRT, and ONNX
• Strong knowledge of performance benchmarking and profiling tools such as Nsight or nvprof
• Proven experience with Infrastructure as Code (Terraform, Helm Charts, or equivalent) and CI/CD pipeline design for GPU-enabled applications
• Working knowledge of Kubernetes and GPU scheduling in containerised environments
• Familiarity with monitoring and observability tools such as Prometheus, Grafana, or NVIDIA DCGM
• Proficiency in scripting languages such as Python, Bash, or PowerShell for automation and monitoring
• NVIDIA certification (preferred)
• Excellent problem-solving, analytical, and troubleshooting skills, with the ability to adapt to evolving requirements
• Strong communication skills with experience collaborating in multicultural teams

If you are a GPU Engineer looking to take the next step in your career and want to contribute your expertise to high-impact optimisation projects in a global environment, this role could be the perfect fit for you. Apply now, or email Richard Fisher - [e-mail]