Senior AI Engineer

We're building the future of mobility.
At hlpy, we're building much more than a company.
We're building the technology platform that is transforming how mobility services are delivered across Europe.
Every month, our platform orchestrates tens of thousands of real-world operations, connecting drivers, insurers, fleets, OEMs, service providers and business partners through technology.
As we continue to expand internationally, broaden our product offering and integrate new businesses, the complexity of what we do grows with us.
Scaling hlpy isn't simply about increasing revenues.
It's about building an organization capable of executing consistently across countries, products and teams.
To get there, strategy alone isn't enough.
Execution is what makes the difference.
That's why we're looking for a genuinely Senior AI Engineer to take ownership of it. This is the most senior hands-on AI role at hlpy, and it comes with a specific kind of mandate. What exists today is deliberately early: prototypes and first working versions, built at speed to prove that the ideas hold. They do. What they were not yet built for is scale, and our evaluation coverage is thinner than we want it to be. Taking that work from promising to production-grade across four countries is the job.
It is an open brief. You would be arriving early enough to shape the foundations rather than inherit them finished.

How the role will shape up
The split will evolve over time, but roughly:
~60% building, running and proving out our production AI agents.
~25% making HLPY's Engineering organisation faster by embedding AI properly into how we build software.
~15% being the company's go-to person for what AI can — and can't — do.
All three areas matter to us. We want to be upfront that this is not just a coding role : you'll have a broad impact across the company, and helping the wider organisation make the most of AI is a core part of the job.

Key Responsibilities

• Own the agent ecosystem, and prove it works Own the design, delivery, reliability and cost of our production agents end to end. Set the architectural direction: agent patterns, model selection, orchestration, and the trade-offs behind each choice. Build and maintain the APIs and MCP servers that connect our agents to the hlpy product and to third-party systems. Establish the evaluation practice properly: datasets built from thousands of real assistance cases, offline evaluation, regression gates before release, and per-country monitoring in production. Build the automated loop — a quality threshold is breached, an issue is created, an agent proposes an enhancement, it is tested, a human reviews and releases it — so that quality does not depend on anyone remembering to check. Know, at any moment, whether last week's change made things better. Today we cannot always answer that. You will fix it.
• Put AI into how hlpy builds software Own our AI-assisted software development lifecycle. Work is under way; making it real is yours. Implement it, keep it current as the tooling changes underneath you, and roll it out across the tech organisation. Train and enable engineers to use it well — and be honest with them about where it does not help. Build the guardrails and documentation that make it safe to use at speed, with Security and Legal where customer data is involved. Treat this as engineering leverage, not advocacy. Done properly it makes every engineer here measurably faster, which is more impact than you will have with any single agent.
• Be the company's AI reference point Be the person teams come to before they buy a tool or start an experiment. Build reusable blueprints and internal documentation so teams can build on your work without you in the room. Run occasional workshops and demos for non-technical teams. Make capabilities feel intuitive to colleagues who do not think in software, and measure yourself by whether they actually use what you built. Tell people plainly when AI is the wrong answer. This is most of the value of having an internal expert. Shape our AI usage guidelines with Security and Legal, so the company can move fast without exposing customer data or drifting into shadow AI.

What This Role Is Not
Not a greenfield build. There is existing work to build on, and part of the job is deciding what to carry forward and what to rebuild for scale.
Not a research position. We ship into production, and production is somebody's bad day at the side of a road.
Not purely hands-on-keyboard. A meaningful share of your week goes on making other engineers better, and that is part of the role rather than a tax on it.
Not a role with a team to delegate to. If you scope it, you build it.
Not unsupervised experimentation. Quality, reliability and data protection come first.

Required

Skills & Experience
Have around 5+ years of professional software engineering experience, including meaningful time building LLM-based systems that real users depend on.
Have taken an agent or LLM feature to production and then operated it — you know what breaks, and why.
Treat evaluation as engineering rather than as a demo. You have built the datasets, harnesses and monitoring that tell you whether a change was an improvement, and you have an opinion about what makes them good.
Design and consume APIs and integrations confidently and independently — REST, and ideally MCP. No specific agent framework is required; we would rather you told us which you would pick here and why.
Are comfortable with the production basics: containerisation, CI/CD, observability, cloud.
Work with real ownership and accountability. You find the problem that matters, scope it, propose the plan, and then deliver it without being managed through it.
Have taken something from prototype to production-grade, and can talk about what had to change.
Hold strong opinions loosely. You argue your case forcefully, and you change position on evidence — including about your own work.
Can explain the same decision to an engineer and to an operations manager without losing precision or talking down to either.

Nice to Have
Scaled an LLM system from early prototype to something a business could depend on, without stopping that business while you did it. This is close to exactly what we are asking.
Introduced AI-assisted development practices to an engineering team and got them genuinely adopted, not merely announced.
Worked in operations-heavy, real-time or multi-country businesses.
Handled AI security and data-protection questions in a customer-data-heavy environment.
A public trail we can look at: repositories, writing, talks.

Representative projects
Concretely, a first year here might contain:
Build the evaluation set we are missing — thousands of real assistance cases, labelled, per country — and use it to settle an architectural question with evidence instead of opinion.
Take a prototype agent to a properly orchestrated production service, and be able to say by how much it improved.
Stand up per-country accuracy monitoring, so we hear about drift from a dashboard rather than from Operations.
Build an MCP server exposing hlpy case and dispatch data, so every new agent stops reimplementing the same lookups.
Take our AI-assisted development lifecycle from work-in-progress to the default way hlpy engineers work: the tooling, the guardrails, the documentation, and the training that makes it stick.
Say no to something. Part of the value of a senior engineer is deciding what not to build, and being able to explain why.

What We Offer
Salary package ranging from € 55.000 to € 65.000 and MBO.
Fresh fruit and good snacks to share with your nice colleagues
Training budget
Remote Working

Why hlpy
Because we're building something meaningful.
Because we move fast.
Because responsibility grows faster than hierarchy.
Because we believe the best ideas can come from anyone.
Because we value ownership over politics and execution over bureaucracy.
And because if we do our job well, millions of people across Europe will experience mobility in a smarter, faster and more reliable way.
If building rather than maintaining excites you, we'd love to meet you.