Semantic data models
The entities, relationships and definitions your domain actually uses, written down so systems can share them.
Services
We build enterprise knowledge graphs that connect data scattered across systems into one structured, source-traceable model. The graph supports search and analytics, feeds graph machine learning and gives AI systems connected context to reason over. Our approach combines semantic modelling, graph reasoning and prediction.
Backed by peer-reviewed research: our ReviewGraph paper was presented at IEEE ICKG 2025. See ReviewGraph research →
The entities, relationships and definitions your domain actually uses, written down so systems can share them.
Pipelines that bring records, documents and sensor data together with consistent identities and provenance.
Turning text and semi-structured sources into entities and relationships, with the source kept for each one.
Embeddings, link prediction and pattern discovery on top of the graph.
Ways to query, explore and visualise the graph, and to feed it into other AI systems.
Give assets, customers or documents one consistent identity across systems, so questions that span them can be answered.
Extract entities and relationships from documents or reviews into a graph and keep the source of each. ReviewGraph does this for customer reviews.
Use graph embeddings and relationship features in machine learning, for example to predict ratings or find missing links.
Keep known evidence separate from model predictions, as in the biomedical graphs of Vink Drug.
Give an LLM or agent connected, source-traceable facts instead of isolated documents.
Agree the entities, relationships and definitions that matter, starting from the questions the graph must answer.
Bring records, documents and sensor data into the graph with consistent identities, extracting entities and relationships from text where needed, and keep the provenance.
Run graph analytics and graph machine learning such as embeddings and link prediction, and validate them against outcomes you already know.
Make the graph queryable through interfaces and APIs, and keep it current as the data changes.
Every project follows the same four steps. Understand → Build → Validate → Deploy. About our way of working →
Research
A peer-reviewed framework that turns customer reviews into a knowledge graph and predicts review ratings from graph embeddings and sentiment.
Read the ReviewGraph research →Research
Biomedical knowledge graphs that keep known evidence apart from model predictions.
See Vink Drug ↗Research
Our current design for a model that infers how an unseen graph behaves from its structure. A vision, not yet a result.
Read the vision →A knowledge graph stores entities, such as assets, people or documents, and the typed relationships between them, together with where each fact came from. Unlike a table, it makes connections explicit, so you can follow them, query them and use them as input for analytics, machine learning and AI systems.
A vector database stores numeric embeddings and finds items that are similar. A knowledge graph stores explicit entities and named relationships, so it can answer questions that depend on how things are connected and can show its sources. They complement each other: embeddings can be computed from a graph, as in our ReviewGraph research.
Yes. A graph can give an LLM or agent connected, source-traceable facts to work from instead of isolated text passages, an approach often called GraphRAG. We describe how we approach this on our AI context layer page.
When several systems or teams use different definitions for the same things, or when answers must be consistent and explainable. You rarely need a large ontology up front. We start with the entities and relationships behind the questions you want answered and extend the model from there.
Yes. We start from the data you already have, such as databases, documents and sensor data, and keep the source of each fact. We only use data that you provide or that is publicly available, under agreed terms.