Founding AI Infrastructure Engineer

Imperial College London; Greater London; On-site; GBP 90,000 - 130,000; Full time

Benefits offered by this job

Equity

Founding-team influence

Small team culture

Job summary

Imperial College London in the United Kingdom invites applications from an exceptionally skilled engineer to build and own our AI training infrastructure. You will work across research and production environments to design scalable systems that run on hundreds of GPUs across cloud providers, emphasizing performance, reliability, and maintainability.

You will collaborate with researchers and platform engineers to optimize throughput, memory management and networking while advancing MLOps tooling

Qualifications

• Experience with large-scale distributed training and PyTorch.
• Strong background in HPC, GPU clusters and memory management.
• Proficient Python and production-grade software engineering.
• Experience with MLOps tooling and CI/CD.
• Ability to work across research and production environments.
• Familiarity with cloud platforms (AWS) and GPU cloud providers.
• Experience with computer vision or multimodal AI is a plus.

Responsibilities

• Design and own large-scale AI training infrastructure.
• Scale distributed model training across hundreds of GPUs and multiple clouds.
• Optimize training throughput and GPU utilization.

Skills

PyTorch

Distributed training

Kubernetes

Slurm

AWS

HPC

MLOps tooling

Python

GPU orchestration

Tools

Kubernetes

Slurm

GPU orchestration

AWS

Job description

What we're looking for:

You'll likely come from an MLOps, AI infrastructure, platform engineering or HPC background, but titles matter less than experience.

• Large-scale distributed training.
• PyTorch and modern deep learning frameworks.
• Kubernetes, Slurm or GPU orchestration platforms.
• AWS and specialist GPU cloud providers.
• High-performance computing and distributed systems.
• Training optimisation, memory management and networking.
• MLOps tooling, CI/CD and experimentation frameworks.
• Python and production-grade software engineering.
• Computer vision, multimodal AI or transformer architectures.

Bonus points if you've worked on:

• Foundation models.
• Research infrastructure.
• Simulation systems.
• Robotics or embodied AI.

Who you are:

You're deeply technical and happiest when solving difficult engineering problems. You thrive in small, ambitious teams. You enjoy building from scratch rather than maintaining legacy systems. You can move comfortably between research and production. You care about performance, elegance and impact. You want ownership, autonomy and the opportunity to shape something significant.

Why join us?

• Join one of Europe's most exciting AI and robotics startups.
• Work alongside leading researchers from Imperial College London.
• Meaningful equity and genuine founding-team influence.
• Solve problems that sit at the cutting edge of AI infrastructure.
• Build technology with the potential to redefine how humans interact with machines.
• We're intentionally keeping the team small, talent-dense and highly collaborative.
• If you want to spend your days solving genuinely difficult problems with exceptional people, we'd love to hear from you.

Robotics is about to have its ChatGPT moment. We're building a new kind of AI system that allows robots to learn complex tasks from a single demonstration. No months of training data collection. No painstaking programming. Show the robot once and it gets to work. Born out of years of research at Imperial College London and backed by one of the largest robotics seed rounds in UK history, we're assembling a small, world-class team to tackle some of the hardest engineering problems in AI. We're looking for an exceptional engineer to own the infrastructure powering our model training. This isn't traditional DevOps. It isn't conventional MLOps. It's a rare opportunity to sit at the intersection of machine learning, distributed systems, cloud infrastructure and robotics.

What you'll do:

Design and own our large-scale AI training infrastructure. Scale distributed model training across hundreds of GPUs and multiple cloud environments. Optimise training throughput, GPU utilisation and