The posting
Architect, build, and scale AI systems that hold up in production
Take ownership of how AI is architected, built, and scaled across our products. You will design the agentic systems at the core of our pension and advisory platforms, with a focus on reliable and responsible AI.
About the role
We are looking for a Senior AI Engineer to help lead our newly established AI team, working on our Life product for pension companies. The team is focused on the practical and responsible use of AI and is still in its early stages, so as a senior member you will play a defining role in setting the technical foundation, design patterns, and ways of working that the rest of the team builds on.
You will work closely with the AI Team Lead and Chief Architect, who own the overall architecture, helping shape the design within that direction. You will also be part of an agile team with a flat structure, where collaboration, curiosity, and continuous learning are central to the culture.
This role is primarily about designing and building agentic and LLM systems end to end, with a meaningful share of your time spent owning how we evaluate them and prove they hold up. It suits someone with 5+ years of relevant experience who wants to move beyond implementation into designing AI systems, and who enjoys raising the bar for the engineers around them.
Why this matters
This is core infrastructure, not edge tooling. Our Life product runs policy and capital administration for some of Europe’s largest pension providers, including PensionDanmark and APG, and is on track to support more than 8 million members and over €800bn in retirement assets.
The AI you design must meet that bar: fast and innovative, but also accurate, secure, and reliable enough for systems that people depend on for their financial security in retirement. Building it well and being able to prove it works, are two sides of the same job.
What you’ll do
As a senior member of the AI team, you will help shape how we build with Large Language Models and agentic systems across our products, contributing to high‑level design decisions, turning ambiguous business problems into robust system designs, and ensuring high‑quality coding standards and design. You will be hands‑on, with a real say in the architecture, standards, and technical judgement that guide the team.
In practice, you will:
Help shape the technical direction
- Contribute to the AI tech stack across our products: help define the high‑level system design, reference architectures, and design patterns the team builds on.
- Help shape AI strategy: identify where AI can create genuine value, frame the trade‑offs, and feed into a coherent, prioritised technical roadmap that fits our wider platform.
- Inform major architectural decisions: model and framework selection, build‑versus‑buy, and how AI components fit into our existing Java platform, integrate with surrounding systems, scale, and stay maintainable over time.
- Reason about trade‑offs in the open: weigh latency, cost, accuracy, security, and reliability, and explain and defend those choices to both technical and non‑technical stakeholders.
Architect and build (hands‑on)
- Design LLM systems end to end: from retrieval, agent orchestration, and data pipelines through to deployment and monitoring.
- Design and build agentic systems: single‑ and multi‑agent, covering orchestration, tool use, action groups, knowledge‑base integrations, and patterns such as routing, delegation, parallelisation, and result aggregation for complex, reliable production workflows.
- Engineer RAG pipelines that connect agents to internal structured and unstructured data.
- Drive quality through design: prompt strategy, system prompts, and memory systems that produce hallucination‑resistant outputs with strong relevance and faithfulness.
- Make cost‑aware architectural choices: balancing latency, cost, and quality across model selection and orchestration patterns so what we ship is sustainable at scale.
- Set the standards for prompt strategy, RAG, orchestration, and cost management that the wider team adopts.
Evaluate and prove
- Build evaluation frameworks: eval sets, golden datasets, and LLM‑as‑judge pipelines, with offline regression suites wired into CI and online evaluation on production traces.
- Evaluate the whole agent, not just the answer: task success, trajectory and tool‑use correctness, and RAG faithfulness and grounding, so you know whether failures come from retrieval, reasoning, or generation.
- Measure what gates shipping: relevance, faithfulness, latency, cost, and reliability across repeated runs, turned into quality gates the team ships against.
Collaborate and grow the team
- Collaborate and mentor: work with the AI Team Lead, Chief Architect, senior developers, and product teams to translate business needs into working, well‑tested solutions, and support more junior engineers as the team grows.
What we’re looking for
You bring a strong technical foundation, hands‑on depth with modern AI tooling, and the judgement to design systems that hold up in the real world. You don’t need deep expertise in every area below, but we do expect real depth in agentic systems and AI architecture, backed by a solid grasp of how to evaluate what you build.
Required
- A master’s degree in computer science or a similar field, preferably with an AI focus.
- 5+ years building software, with significant hands‑on work applying Large Language Models.
- Strong proficiency in a statically typed, object‑oriented language, Java or C# (we build in Java). Python is welcome for demos, evaluation tooling, and experiments, but it is not our production language.
- Deep experience designing, building, and operating agent‑based systems, with hands‑on use of one or more orchestration frameworks (e.g. LangChain/LangGraph, Microsoft Agents, Claude Agent SDK, AutoGen, Semantic Kernel).
- A track record of designing and building AI systems, including how the components fit together, scale, and are maintained over time.
- Solid, practical experience with prompt strategy, Retrieval Augmented Generation (RAG), and the cost considerations of LLM usage and model selection.
- Working knowledge of LLM and agent evaluation: eval sets, LLM‑as‑judge, and the metrics that distinguish a good system from a confident‑but‑wrong one (faithfulness, relevance, task success, latency, and cost).
- Experience designing AI or backend systems that fit into a larger production platform, collaborating with senior engineers and architects to align on a shared technical direction.
- Strong communication skills to explain and defend architectural trade‑offs to both technical and non‑technical stakeholders.
Nice to have
- Experience building automated evaluation pipelines or CI gates for AI systems.
- Familiarity with observability and tracing tooling for LLM applications (e.g. LangSmith, Langfuse, Arize Phoenix, Braintrust).
- Familiarity with LLM evaluation, automation, and red teaming.
- Broader machine learning experience.
- Hands‑on experience building and integrating MCP servers.
- Experience working in regulated or high‑reliability domains such as finance, insurance, or healthcare.
Who you are
You are a confident, hands‑on engineer who takes ownership of complex and sometimes undefined problems, and who is just as comfortable designing a system as proving it holds up. You think in systems, balance pragmatism with quality, and care about building things that work in production rather than just in a demo. You lift the people around you, stay open to feedback, and bring strong technical judgement of your own. You are curious, persistent, and result‑oriented.
