MS

Experienced AI Engineer

Hiring from
Denmark
Work type
Hybrid
Posted

625,831 remote jobs, straight from company career pages

100% free · New jobs every minute

Show job description

About MicroTech Software

MicroTech is a small software company with more than 35 years of experience building custom solutions in close dialogue with our customers, mainly in banking and insurance. Since 2025, applying AI has been a major focus for us.


We value quality over quantity, and we keep the customer experience at the centre. Expect great colleagues and a busy but relaxed workday in an informal atmosphere. We also care about ESG and like to give back, for example through sponsorships and by being an inclusive workplace that welcomes employees who need special accommodations.


About the role

You’ll help shape the architecture of a live AI product running on our own on-prem LLM stack.


Two big projects are coming up, and you’ll play a key part in both:

  • Voice-to-workflow: making speech a first-class input channel in our agentic workflows
  • Moving to a graph architecture: going from RAG to GraphRAG by building a knowledge graph and integrating it into our existing pipeline


Responsibilities

Shape the architecture of a live AI product

  • Evaluate and challenge design decisions across the stack, including retrieval strategies, LLM backend selection, service boundaries and data flow, and propose improvements with clear reasoning
  • Take part in technical planning with the team and with insurance-domain stakeholders, turning vague requirements into concrete engineering tasks
  • Identify technical debt, assess its cost and make the case for addressing it


Develop and optimize our on-prem AI stack

  • Further develop and optimize our self-hosted LLM stack: local models, local embedding and on-prem vector stores
  • Evaluate open-weight LLMs for Danish-language RAG quality, weighing latency, accuracy and hardware requirements


Move to a graph architecture (RAG → GraphRAG)

  • Help design and build a knowledge graph for our domain, including ontology and knowledge extraction
  • Integrate the knowledge graph into our existing RAG pipeline and measure the effect on retrieval quality


Design and build agentic workflows

  • Design multi-step agent architectures using LangGraph or equivalent frameworks that can route, plan, use tools and synthesize across document corpora
  • Define tool-use interfaces, memory patterns and failure-handling strategies for agent loops running in production
  • Work with domain experts to identify automation opportunities worth building


Enable voice-to-workflow capabilities

  • Help design a pipeline that integrates local speech-to-text models into agentic workflows
  • Evaluate Danish-language ASR quality, identify gaps and assess whether fine-tuning or model selection is the better approach
  • Design the integration points carefully, with attention to latency, error handling and data privacy, so voice becomes a first-class input channel


Requirements

  • 2–4 years of production Python engineering experience, including async code, clean architecture patterns and testable codebases
  • Direct experience with LLM orchestration frameworks (LangChain or similar) in a shipped product or serious project
  • Solid understanding of RAG systems: chunking strategies, embedding models, vector search and retrieval quality evaluation
  • Familiarity with vector databases such as pgvector, Chroma, Weaviate or Pinecone
  • Comfortable owning a FastAPI service end-to-end: schema design, error handling, auth,
  • containerization and deployment
  • Ability to reason about architectural trade-offs and communicate them clearly
  • Hands-on experience designing or running local LLM inference (Ollama, vLLM, llama.cpp) in a real rather than toy context


Nice to have

  • Familiarity with knowledge graphs, including ontology building, knowledge extraction and RAG integration (especially relevant)
  • Speech-to-text pipeline work (Whisper, Deepgram, Speechmatics or comparable local STT), especially on non-English languages (especially relevant)
  • Experience with agentic frameworks (LangGraph, AutoGen, CrewAI) or custom tool-use and planning loops in production
  • Hybrid retrieval systems (dense + sparse, BM25 + vector) and re-ranking with cross-encoders
  • LLM observability and tracing tools (Arize Phoenix, LangSmith or similar)
  • Multilingual NLP, particularly Danish or other Nordic languages
  • Knowledge of regulated-industry data handling, GDPR or EU AI Act compliance
  • Cloud platforms (e.g. Azure), container registries and CI/CD


What we offer

  • Beautiful offices: art on the walls, plenty of green plants and high ceilings
  • Great location: bus, train and light rail right outside the door, easy to reach by car, and free parking
  • Flexible work: a hybrid setup with the option to work from home up to around 2 days a week after the initial onboarding
  • Small organisation: very short decision-making processes and great opportunities for co-responsibility
  • Personal development: varied tasks, direct client contact and support for further education
  • All phases: you’ll work on everything from definition and solution design to programming and production
  • Competitive salary: an attractive compensation package that matches your skills


Please note

  • Start date: as soon as possible. We’ll still take the time to find the right person, so we review applications on an ongoing basis and will keep the position open until it’s filled.
  • This is a hybrid role based at our office in Viby J., Aarhus. It is not a fully remote position.
  • We cannot offer work permit sponsorship or relocation support. Applicants must already have the right to work in Denmark.


Similar jobs

Apply for this job100% free