UDA Addendum: Giving Thanks

Jeff Uren ★ 2026-08-14

This is an addendum, or follow-up to the UDA article series on this site, which you can find here:

Before finishing this series, I wanted to take a moment to acknowledge something that is easy to lose sight of when talking about architecture.

Almost none of these ideas began with us.

Like most engineering, what we're building stands on the shoulders of an extraordinary number of people who spent decades exploring problems that many organisations are only now beginning to encounter.

The Upper Data Architecture, DataMesh, the query engine, the generated projections, the semantic APIs, the integration spine, and many of the surrounding ideas are our implementation of a much broader body of work. They represent a year of reading papers, building prototypes, throwing ideas away, arguing, and learning from organisations that were generous enough to publish what they had discovered and in some instances, lending us their own time to help us form our own understanding.

In particular I'd like to thank the engineers, architects, and researchers at organisations like Netflix, NASA, Bosch, NHS, and many others who have consistently chosen to publish their work rather than keep it hidden behind organisational walls.

Many of the ideas explored throughout this series (knowledge graphs, semantic modeling, metadata-driven systems, ontology engineering, graph-based reasoning, data lineage, federated data platforms, schema evolution, and executable metadata) exist today because those communities were willing to document both their successes and their failures.

Perhaps more importantly, the academic community has spent decades laying the theoretical foundations that make systems like this possible. Work around ontologies, RDF, OWL, SKOS, semantic interoperability, knowledge representation, and linked data has often been viewed as niche or overly academic. Increasingly, however, those ideas are proving remarkably practical as organisations struggle to make sense of a rapidly growing and increasingly inter-connected data landscapes.

The implementation described throughout these articles should therefore be viewed for what it really is: another step and some synthesis in a much longer journey.

Our hope was never to to invent an entirely new way of thinking about data.

Our hope is to take the remarkable work that has already been done, combine it with the practical realities of building and operating a modern data platform, and make those ideas useful within Wellcome, and hopefully, in the future, the broader funding and research community through example.

If we succeed, the real beneficiaries won't be the platform team.

They'll be the analysts who spend less time hunting for data, the engineers who spend less time rebuilding integrations, the domain experts who can finally describe their part of the organisation in a way that software can understand, the governance teams who gain clearer visibility into how data moves, the researchers who receive answers more quickly and with greater confidence, and ultimately, the scientific work that Wellcome exists to support.

There's still an enormous amount to build, there are undoubtedly ideas in this series that will prove naive, others that will need refining, and some that may turn out entirely wrong.

That's part of engineering though.

But if these articles encourage even a handful of people to think about organisational meaning before storage, concepts before schemas, and semantics before implementation, then they've achieved exactly what I hoped they would.

FAIR, data mesh, and data platform governance

The FAIR Guiding Principles for Scientific Data Management and Stewardship — Mark D. Wilkinson et al.
Data Mesh: Concepts and Principles of a Paradigm Shift in Data Architectures — Ivo A. Machado et al.
Data Mesh: A Systematic Gray Literature Review — Abel Goedegebuure et al.
Ten Pillars for Data Meshes — Robert L. Grossman et al.

Biomedical / science-domain semantic platforms

The EBI RDF Platform: Linked Open Data for the Life Sciences — Simon Jupp et al.
Open Targets: A Platform for Therapeutic Target Identification and Validation — Gabriela Koscielny et al.
The Next-Generation Open Targets Platform: Reimagined, Redesigned, Rebuilt — David Ochoa et al.
Biolink Model: A Universal Schema for Knowledge Graphs in Clinical, Biomedical, and Translational Science — Deepak R. Unni et al.
KG-Hub: Building and Exchanging Biological Knowledge Graphs — J. Harry Caufield et al.
Announcing the Biomedical Data Translator: Initial Public Release — Karamarie Fecho et al.

AI / knowledge graph grounding

Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks — Patrick Lewis et al.
From Local to Global: A Graph RAG Approach to Query-Focused Summarization — Darren Edge et al.
Unifying Large Language Models and Knowledge Graphs: A Roadmap — Shirui Pan et al.
A Survey of Graph Retrieval-Augmented Generation for Customized Large Language Models — Qinggang Zhang et al.

UDA Addendum: Frequently Asked Questions

UDA Addendum: Giving Thanks

Suggested Reading

Foundations of the Semantic Web

RDF / OWL / SKOS

Knowledge Graphs

Enterprise Knowledge Graphs

Metadata & Data Platforms

data Lineage & Metadata

Query Engines

W3C RDF / OWL & Ontologies

General Concepts

The Bosch Papers

Core ontology / semantic data platform papers

Knowledge graphs and semantic data management

FAIR, data mesh, and data platform governance

Biomedical / science-domain semantic platforms

AI / knowledge graph grounding