Junior AI / Machine Learning Engineer

Junior AI / Machine Learning Engineer

SpaceAM · On-site · Full-time · UK Passport holder

Machine learning that runs on a chip the size of a fingernail

Most machine learning gets to be lazy. It runs in a data centre with effectively unlimited memory and power. Ours doesn't. We build ML models that run on the device: real-time inference on tiny, low-power embedded hardware, with no cloud, no GPU, and only a few hundred kilobytes of memory to work with. Getting a model to be both accurate and small enough to live on a microcontroller is a genuinely hard, genuinely interesting engineering problem, and it's the one you'd spend your days on.

Our ML function is small and growing. This role puts you alongside our lead ML engineer from day one, working across the whole pipeline: training the models, running the infrastructure that makes our experiments reproducible, and getting models onto real hardware. You'll have real ownership early and a clear path to becoming an independent engineer who can carry work end to end.

The work supports products in the security, defence, space and medical sectors. We'll talk through the specifics with you as the process progresses.

What you'll do

· Train and evaluate models. Run experiments on our models (built in PyTorch), analyse the results, and figure out why a model behaves the way it does, not just which knob to turn next.

· Do evaluation honestly. Design experiments that give real answers: proper data splits, multiple runs, the right metric, and the judgment to know when a result is signal and when it's noise.

· Work with data. Ingest, organise, version, and prepare the datasets our models learn from.

· Keep the ML infrastructure running. Look after the experiment-tracking and data-versioning stack that underpins everything we do, run training jobs across our compute, and debug the failures that come with real systems.

· Grow toward the edge. Over time, learn how a trained model becomes a compressed model that runs on a microcontroller, and help prove that what runs on the device matches what we trained. This is a deliberate growth area; the ceiling is you owning the full path from model to hardware.

You won't do all of this on day one. You'll start where you're strongest and grow into the rest with support.

What we're looking for...

You don't need to tick every box. We care far more about how you think than about which tools you've already used.

We think you'll need:

· Solid Python. You're comfortable working in a real codebase, not only notebooks.

· A genuine grasp of machine-learning fundamentals: training vs. validation, overfitting, class imbalance, and how to read a metric critically.

· Comfort on the Linux command line and with git.

· Rigour and honesty. You want to understand what's happening rather than copy-paste until it works, and you'll happily say "I'm not sure, let me check."

· The ability to teach yourself a new tool from its documentation.

Any of these would be a bonus (none are required):

· Hands-on PyTorch or another deep-learning framework.

· Signal-processing background.

· Any MLOps tooling: experiment trackers, data-versioning tools, or object storage.

· Docker or other containers.

· Embedded / TinyML exposure: microcontrollers, C/C++, or model quantization.

· A sound feel for experiment design and statistics.

Qualifications

· A degree in a technical or quantitative discipline (Computer Science, ML/AI, Data Science, Engineering, Physics, Maths, or similar), or equivalent ability shown through projects, work, or self-study. We assess the ability, not the certificate.

· Some demonstrable hands-on ML: projects, internships, competitions, open source, or prior experience. We want to see you've built and evaluated a model, not only studied it.

· This is a junior role (roughly 1–2 years' experience). We're not expecting years of experience or a PhD. We're looking for strong instincts and a drive to learn fast. Strong graduates and career-changers welcome.

Eligibility & clearance

This role involves work that requires security clearance. You must be eligible to obtain SC clearance. We're happy to answer questions about this early in the process.

What we offer

· 35k to 45k

· We operate a late start scheme depending on travel time.

· Pension Scheme

· Cycle to Work Scheme

· Free parking

· 5 weeks annual leave, plus bank holidays

· The chance to work on hard ML problems that most engineers never touch, with real ownership and a clear path to grow.

· Annual leave and paid sickness entitlements increase with the length of service.