About the position
Join MicroTech Software to build cutting-edge AI-powered tools for insurance professionals. Work on production RAG chatbots, agentic workflows, and voice-to-workflow capabilities.
At MicroTech Software we build cutting edge AI-powered tools that help insurance professionals answer complex questions faster and with greater confidence. Our flagship product is a production RAG chatbot used daily by insurance workers, and we are actively pushing its architecture in several new directions at once using the newest models. Your reference will be our AI Lead. We are looking for an engineer with 2–4 years of experience who has reached the point where writing good code is no longer the hard part — and who is ready to take on the harder questions: which approach to use, how to structure a system for long-term maintainability, and where to invest engineering time to get the most impact. You will work in a team with a very advanced team lead whom you can expect to learn from and alongside one or more junior engineers and be expected to help set the direction of their work as well as your own.
Responsibilities
- Shape the architecture of a live AI product
- Evaluate and challenge design decisions across the stack — retrieval strategies, LLM backend selection, service boundaries, data flow — and propose improvements with clear reasoning
- Participate in technical planning with the team and with insurance-domain stakeholders, translating vague requirements into concrete engineering tasks
- Identify technical debt, assess its cost, and make the case for addressing it
- Navigate an on-prem transition
- Design and lead the migration of a cloud-dependent LLM stack to a fully local alternative: self-hosted models, local embedding, and on-prem vector stores
- Evaluate open-weight LLMs for Danish-language RAG quality; understand the trade-offs in latency, accuracy, and hardware requirements
- Contribute to a RAG → GraphRAG transition: build a knowledge graph and integrate it into an existing RAG pipeline
- 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
- Contribute to the design of a pipeline that integrates local speech-to-text models into agentic workflows
- Evaluate Danish-language ASR quality, identify gaps, and assess the feasibility of fine-tuning or model selection approaches
- Design the integration points carefully — latency, error handling, data privacy — so voice becomes a first-class input channel
We are looking for someone who can say yes to
Must-have
- 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, retrieval quality evaluation
- Familiarity with vector databases — pgvector, Chroma, Weaviate, Pinecone, or equivalent
- Comfortable owning a FastAPI service end-to-end: schema design, error handling, auth, containerization, deployment
- Ability to reason about architectural trade-offs and communicate them clearly
Strong advantages
- Hands-on experience designing or running local LLM inference — Ollama, vLLM, llama.cpp — in a real rather than toy context
- Familiarity with knowledge graphs — ontology building, knowledge extraction, RAG integration
- Experience with agentic frameworks: LangGraph, AutoGen, CrewAI, or custom tool-use / planning loops in production
- Speech-to-text pipeline work — Whisper, Deepgram, Speechmatics, or comparable local STT — especially on non-English languages
- Experience designing hybrid retrieval systems (dense + sparse, BM25 + vector) and re-ranking with cross-encoders
- Familiarity with LLM observability and tracing tools (Arize Phoenix, LangSmith, or similar)
- Exposure to multilingual NLP, particularly Danish or other Nordic languages
Nice to have
- Knowledge of regulated-industry data handling, GDPR, or EU AI Act compliance considerations
- Familiarity with cloud platforms (Azure preferred): managed services, container registries, CI/CD
Interested?
Send your application via our online form or directly to us.
Or send to [email protected] · Call us at +45 7585 9444 for questions
We offer
- • Beautiful offices at Viby J. station
- • Free parking
- • Opportunity to work hybrid
- • Competitive salary
- • Short decision-making process
- • Personal development