ML Platform Engineer

ML Platform Engineer | Python, model serving, distributed systems | Fully remote, UK
£90,000 to £115,000. Permanent.
London. Fully remote across the UK.

The models are only useful if the team can train, evaluate, deploy and observe them without rebuilding the machinery each time. You will own that machinery.

Our client is an early-stage AI product company building applications that get on with everyday tasks before you ask. A prototype exists, launch is ahead, and the company is funded without relying on an upcoming round.

You’ll work across the infrastructure behind the AI stack, alongside AI engineers, researchers and product engineers. The aim is to make experimentation quick, production releases dependable and the expensive parts of the stack visible.

This is infrastructure work with ML consequences. Throughput, latency, GPU use, cost, model regressions and failed releases all land here. You need to be comfortable debugging across the line between platform and model serving.

The job
Platform. Build and operate the systems used for model training, evaluation, deployment, inference and experimentation. The useful outcome is reusable infrastructure that other engineers can rely on, not a pile of one-off pipelines.
Serving. Build and optimise inference infrastructure for high-throughput, low-latency workloads. You’ll work with serving technologies such as vLLM, SGLang or TensorRT-LLM and find the bottlenecks across GPU and distributed systems.
Pipelines. Develop repeatable flows for data preparation, training, evaluation, model release and continuous improvement. Reproducibility and maintainability matter as much as getting a first run through.
Observability. Build monitoring, tracing, alerting and benchmarking for AI workloads. You’ll make model and infrastructure regressions visible early enough to do something about them.

What you’ll bring
Essential:

• ML infrastructure or production ML platforms you have built or operated.
• Production Python used for platform or machine learning systems.
• Model deployment, inference or evaluation infrastructure you have owned.
• Distributed systems work where reliability mattered in production.
• Low-latency or high-throughput serving you have tuned.
• ML or data pipelines built to be repeatable and observable.
• Production monitoring, tracing or alerting for ML workloads.
• Existing right to work in the UK.

Useful: PyTorch, JAX, vLLM, SGLang, TensorRT-LLM, GPU tooling, vector databases, workflow orchestration.

Who this suits
You are probably an ML Platform Engineer, MLOps Engineer or Machine Learning Infrastructure Engineer who likes building the common layer underneath model teams. You think about the whole route from experiment to serving.

It won’t suit you if you want to run infrastructure and leave the models to someone else. Here a slow endpoint or a regression after a release is yours to trace, through the GPUs, the serving layer and the model itself.

Let’s talk
Actively looking or just curious, apply through Xist4 for a confidential conversation. We welcome applicants from every background and will support reasonable adjustments.