Applied AI Engineer

Tech-first. Ego-free. User-obsessed.

Over one million Danes use Skatteguiden to understand and optimise their taxes. We automate tax tracking, surface deductions people didn't know they had, and give users the insights they need to make better financial decisions — all directly connected to the Danish tax authority.

We're an independent company of around 50 people based in Copenhagen, with a 4.7-star rating and a simple mission: make financial insight accessible to everyone, not just those who can afford an accountant.

We have an ambitious multi-year vision for where we take the product next. This role helps build the conversational and agentic layer that makes it credible.

About the role

You'll work primarily on the LLM-powered parts of Skatteguiden. Today, that's a mix of customer-facing and internal products: RAG-based assistants and agentic systems, running on Azure infrastructure.

Over time, the agentic side of this stack will grow significantly, and the infrastructure needed to get there safely is part of the job.

This is a role about building agentic systems the right way — with quality, evaluations, and guardrails as first-class concerns. Alongside your team, you'll regularly partner with our Head of Security & Privacy and our Legal/Compliance team.

You'll join the AI Products team, working closely with our existing Data Scientist, our AI Engineer, and our incoming Senior ML Engineer. While we each have our specialisations and focus, we enjoy collaborating across our roles as our team is small and dynamic.

What you'll do

- Build on our existing LLM products. Ship improvements to quality, latency, cost, and safety. These are production systems used every day by users and internal teams.

- Build out the evaluation infrastructure. Offline evals, online validation, regression tracking, LLM-as-a-judge pipelines where they make sense. Evals aren't an afterthought at Skatteguiden; they drive our product decisions.

- Help define SLOs. Latency, quality, cost, and safety. Formal SLOs for production LLM systems are part of the bread and butter. You'll help define them.

- Contribute to the agentic trust infrastructure. Human-in-the-loop review flows, explicit consent, reversibility, and audit trails. As the product moves towards further maturity, this is what separates a product users trust from one they don't.

- Partner on red teaming and compliance. Work with Security & Privacy on adversarial testing and with Legal on regulatory fit.

- Stay on top of the model landscape. Azure today, but the landscape moves fast. You'll help us stay current and avoid lock-in.

Who we're looking for

You're an engineer who has built LLM systems and has opinions about how to do it well. You think evals are a first-class concern. You design for challenges like hallucinations and prompt injection. You know agentic systems are exciting, but don't ship themselves.

You've shipped LLM systems that real users touched, you take evaluation seriously as an engineering discipline, and you have the instinct to know when something is working and when it isn't. If that's you, we'd like to hear from you.

This role spans more than just LLM work. Evaluation infrastructure, integration, guardrails, and the critical engineering around production AI systems are a big part of the job.

Must-have

- Working experience building LLM-powered applications in production. Not a weekend demo — something real users have touched.

- Hands-on grasp of RAG, tool use, and at least one orchestration framework (LangGraph, OpenAI Agents SDK, Claude Agent SDK, or equivalent).

- Experience with LLM evaluation; offline evals, online validation, or regression tracking. You treat quality as an engineering discipline, not a gut check.

- Solid software engineering foundations. Python, clean code, testing, CI/CD.

- Security and privacy awareness. You understand why guardrails matter for a product that handles financial data.

- Good judgment under ambiguity. LLM work is full of unclear trade-offs; you can reason about them and pick a direction.

- Genuine enthusiasm for AI. You use AI tools daily in your own work, and you're always looking for new ways to apply them.

Nice-to-have

- Experience with production agentic systems — tool use, planning, human-in-the-loop review flows.

- Experience with Azure OpenAI and Azure AI Search.

- Experience with Agent-to-Agent (A2A) or similar protocols

- Prior work in fintech, healthtech, or another regulated domain.

- Familiarity with LLM observability tooling.

- Danish language is useful but not required. The team works in English.

Why join us

- A vision with real ambition. From conversational guidance today to an AI Companion that helps users plan and eventually act on their behalf. You'll help build the layer that makes it safe and credible.

- Quality and evals as a first-class discipline. Not a checkbox. Not a sprint goal. A permanent part of the job, with real backing from leadership.

- Room to grow. Mid-level today, but the scope of the work and the team's growth mean you'll have real room to step up. This is a good role to grow in seniority.

- The product matters. Over one million Danes use Skatteguiden. Your work keeps it trustworthy.

- Solid company. Independent, profitable, growing. No VC pressure, no bloated roadmap. Just a clear mission and a team that cares.

- A company that's serious about AI. We use AI across the business, and we're only going deeper. Being a small, focused team means we can move fast on this — no lengthy approval chains, no slow rollouts.

- Good people. We take the work seriously without taking ourselves too seriously. The culture is professional, informal, and genuinely collaborative.

Practical details

Location: Copenhagen. We're an office-first company, and it's a deliberate choice. We're moving fast, we have big ambitions, and we genuinely believe that being physically together is one of the things that makes that possible. Things move quicker when you can turn your chair and talk to someone.

We're a five-minute walk from Rådhuspladsen, stocked with everything from fruit to fizzy drinks to afternoon snacks, and lunch is at Claus Meyer's canteen just across the street — proper food, not an afterthought. If you're putting in a long day, dinner is covered too.

Start: As soon as possible.

Salary: Competitive and based on your experience.

Our lean hiring process

1. Screening call with our recruiter (30 min) ONLINE

2. Technical interview with the Head of AI Products and an AI engineer (120 min) ONSITE

3. Case / technical deep-dive with the AI Products team (60 min) ONSITE

4. Final conversation with the CTO and CEO (30 min) ONSITE