About this role
Summary: Own the AI layer of a legaltech platform, including retrieval, orchestration, and evaluation of LLMs for document analysis. Build and maintain production-grade systems for clause extraction, redrafting, and compliance monitoring while collaborating with full-stack engineers.
Contractify and Legile are two Belgian legaltech companies in the same group. Togetherwe build one platform: contract management plus AI that reads and analyses legaldocuments. It runs in production and is used daily by hundreds of customers.
We are looking for an AI/ML Engineer to own the AI layer of that platform: retrieval,orchestration, the prompts and models behind document comparison, clause extraction,redrafting and anonymisation, and the evaluation that tells us whether it works. Much ofthat knowledge sits with external partners today. Your job is to bring it in-house and buildon it.
Why this role
Our users make decisions based on what the system returns. A hallucination is a risk. Amissed clause is a problem for a lawyer.
The foundations are there: multi-tenant setup, RAG, document comparison,anonymisation, a growing customer base. What is missing is one person who understandsevery pipeline end to end, measures quality, and decides when a new model or approachis worth it. You work in one product and development team across Contractify and Legile,with one backlog, next to the full-stack engineers who take your work to production.
What does your day look like?
- Retrieval: how the platform finds the right passages in large volumes of legal text:chunking and embedding strategies, vector and hybrid search that handles cross-references and defined terms. Target: recall above 85% and latency under 500 ms onmore than 10,000 Belgian legal documents.
- Orchestration and agents: the workflows that chain LLM steps for documentcomparison, clause extraction, redrafting and anonymisation, plus a multi-agent setupthat monitors legal and regulatory changes and alerts only when it matters.
- Domain-specific models: build annotated legal datasets with our in-house legalexperts, using as little of their time as possible, and use them to test and adapt modelsfor clause comparison, citation accuracy and multi-turn Q&A. Decide with data whenfine-tuning beats prompting.
- Prompts and models: version and test the prompts and model settings behind each pipeline. Evaluate new models and decide when to switch.
- Evaluation: build the evaluation setup for retrieval quality, hallucination detection and regression testing on real documents, so "is the AI getting better?" has a number behind it.
- Document understanding and integrations: OCR and layout analysis, structured extraction, anonymisation of personal data before it reaches a model, and a plugin architecture that connects the AI layer to several document management systems.
- Predictive compliance (R&D): explore whether contract data, a graph database and time-series models can flag compliance risks early.
- Internal AI agents: we are building agents and automations for our own sales, marketing and customer success teams with an external partner. You are the technical owner on our side: where they run, how they are secured, and how they keep running once the partner steps back.
- Knowledge: take over the AI/ML knowledge from our external partners, document it, and be the person the team turns to for LLM, retrieval and evaluation questions.
- AI-native development: you use AI coding tools daily for coding, requirements and tests. The team is being coached on this; you help build the shared assets (skills,guardrails, agent workflows) and keep them up to date afterwards.
What do we offer you?
- A competitive salary package
- Real R&D, partly funded by Flemish innovation grants (VLAIO)
- A live product with real users: your work is visible from day one
- Ownership of the AI core: your choices on retrieval, orchestration and evaluation
- shape the roadmap
- In-house legal experts who validate what the AI produces
- Hybrid work from our Ghent offices and home work
- Learning budget: conferences, courses, certifications
- Part of a larger technology group, with the network and career options that come with it.
- * *
What are we looking for?
#### Your technical skills
- 3+ years building LLM-based applications, retrieval systems or NLP for real users, with strong Python
- Hands-on RAG experience: chunking, embeddings, vector search (pgvector or similar),
- hybrid search, re-ranking
- An LLM orchestration or agent framework, and a view on when not to use one
- Experience evaluating AI systems: retrieval metrics, hallucination detection, regression sets
- At home in a production codebase: Git, PR reviews, tests, Docker, cloud
- You use AI coding tools daily and can show how they make you faster without lowering quality
#### Your soft skills
- You are pragmatic about AI. You can read a paper on retrieval and have a working version running the same week, then measure whether it helped.
- When someone says "the AI sometimes gets this wrong", you build the test set before youbtouch the prompt. You have opinions on prompts, models and evaluation, and changebthem when the data says so.
- You don't need to be a lawyer, but legal text should interest you rather than put you off.
#### Your degree
- Master's in Computer Science or AI, or equivalent experience
#### Your language skills
- English
- Nederlands