Applied AI Engineer
Confirmed on the employer's own hiring board on Nov 9, 2026. First seen by Alion on May 21, 2026. Seamflow scores B on the Alion truth index.
About Seamflow
Every plane, medical device, lift, battery and building has to be tested, inspected and certified before it reaches the world. It's the invisible layer that keeps the physical economy safe, and it still runs on Word documents, PDFs, spreadsheets and email. Seamflow is the software layer for that industry. We build AI-powered workflows for certification bodies, notified bodies and inspection companies: scheduling and resource planning, technical review, conformity assessment, field inspection, and document intelligence over messy real-world files.
Where we are:
Live in production inside some of the largest enterprises in the industry
$4.5M seed, 14x oversubscribed, co-led by Northzone and Initialized Capital, the funds behind Spotify, Klarna, Coinbase and Reddit. Angels include Charlie Songhurst and Mario Götze.
Demand is outstripping what we can build. That's the reason this role exists.
The team
Most of our engineers were olympiad medallists in maths, physics or informatics before they were engineers. The team has come to us from Google, Amazon, X, Yandex, McKinsey and Bain, and from founding-engineer seats at startups that went on to multi-billion-dollar valuations. We hire against a very high bar, and this role exists to raise it further.
The role
An Applied AI Engineer here thinks in workflows and model behaviour, not benchmarks or isolated prompts. The inputs are genuinely hard: 400-page technical files, scanned forms, tables that were never meant to be parsed, documents where a single misread classification has consequences. The bar is not a good demo. The bar is a system a reviewer trusts on a Tuesday afternoon with a deadline.
You own AI surface area end to end: what the system should do, the prompts, the agent design and tool use, the evals that catch regressions before customers do, the monitoring, and the iteration after it's live. You'll sit in customer calls, read the real documents, and look at hundreds of real outputs, because that's where the failure modes actually are. You'll ship to production in your first week.
What we're looking for
We care far more about raw engineering strength than about matching a list of technologies. If you're excellent, we'll hire you whether you've spent the last three years in Rust, Go, Python, TypeScript, C++ or something we haven't heard of. Our stack is learnable in a fortnight. Judgement, speed and standards are not.
What we're actually screening for:
You've run AI systems in production, not demoed them. Something real users depended on, that broke in ways a benchmark never predicted, and that you fixed.
You can say why a model failed. Not that it hallucinated. Why, and what you changed.
You build evals that catch regressions, rather than metrics that flatter the system.
Speed with taste. You ship fast, and you're opinionated about what AI should not do and when good enough ships.
You go to the source. You'll read the customer's actual documents, join the actual call, and look at hundreds of real outputs.
Signals we like, none of them required:
multi-step agents and orchestration you've run in anger, document intelligence over genuinely messy inputs, founder or ex-founder, founding team experience, EF / YC / Antler, early or stealth-stage startup, competitive programming or olympiad background, open source you genuinely maintain.
This is probably not for you if
You want a strict boundary between work and life. We run one of the most ambitious startup cultures in Europe: faced paced, extremely high ownership, and in person in Shoreditch.
You want a well-defined ticket handed to you each morning.
You want to specialise narrowly and stay in your lane.
What you get
Real equity, and compensation that reflects a high bar
Genuine ownership of product surface area from day one, with the founders in the room
Relocation support and visa sponsorship, handled properly
Lunch and dinner covered, 24/7 gym access, and any tool that makes you faster
A small team with high standards, high trust, and no politics
If this sounds like your kind of problem, apply. A short note on the hardest thing you've built goes further with us than a polished CV.