The project
I just finished a full upgrade of my SKOS-XL search-template application — a semantic-web taxonomy tool I originally built to explore controlled-vocabulary search and concept relationships. The new version is modernized end-to-end with GitHub Copilot, refreshed dependencies, and a tighter architecture that makes it usable as a component in real-world AI systems, not just a standards demo.
If you work in data architecture, AI/ML infrastructure, knowledge engineering, or enterprise search, this is the kind of tooling layer that sits between messy source data and useful AI outcomes.
Why SKOS-XL still matters
SKOS (Simple Knowledge Organization System) is the W3C standard for sharing and linking vocabularies, taxonomies, and thesauri. SKOS-XL extends it with support for labels as first-class resources — essential when you need to track multiple labels per concept, provenance, translations, and editorial history.
In enterprise terms, SKOS-XL is a building block for:
- Controlled vocabularies and metadata tagging
- Taxonomy / ontology management for regulated industries
- Search and findability across data catalogs, document repositories, and product catalogs
- Retrieval-Augmented Generation (RAG) where LLMs retrieve from curated, governed concept graphs rather than raw embeddings alone
- AI governance and auditability — you can explain why a concept was matched
What I upgraded
The project was originally a useful but dated semantic-web experiment. The upgrade focused on making it maintainable, observable, and deployable:
- Refreshed the build and runtime to modern, supported versions
- Rebuilt the front-end search interface with a lightweight, current component model
- Used GitHub Copilot throughout to accelerate refactoring, test coverage, and documentation
- Preserved the SKOS-XL data model while making the API cleaner and more REST-friendly
- Documented the architecture so the project can serve as a reference implementation for taxonomy-backed search
The result is a small but complete stack: a searchable SKOS-XL concept browser, a concept-detail view, and a backend that can load and query real taxonomy data.
Where this fits in AI/ML work
Most production AI projects spend 80% of their effort on data preparation, lineage, and governance. A SKOS-XL layer gives you a governed semantic index that can:
- Normalize terminology across business units and data sources
- Power retrieval for RAG pipelines with explainable, human-curated relationships
- Support data catalogs and metadata management by tagging datasets with standardized concepts
- Improve search relevance in healthcare, finance, pharma, and public-sector systems where precision matters
If you’re building a data platform, AI governance framework, or enterprise knowledge graph, taxonomy work is not a side quest — it is the foundation.
What I’m looking for next
I’m a senior Data Architect / Knowledge Graph / AI Architecture practitioner with a background in regulated enterprise environments (healthcare, federal, insurance). I bring deep experience in:
- Taxonomy, ontology, and metadata framework design
- Enterprise data modeling and semantic data architecture
- Search, data catalog, and lineage systems
- Data governance, quality, and measurement
- Standards-based interoperability (W3C, SKOS, RDF, XML, XQuery, JSON-LD)
I’m exploring senior roles where taxonomy, metadata, and measurement frameworks directly improve AI outcomes — especially in organizations that need AI systems to be accurate, governable, and audit-ready.
If you’re a recruiter, hiring manager, or engineering leader working on AI infrastructure, knowledge graphs, or enterprise data platforms, I’d welcome a conversation.
Try the project
Repo: [https://github.com/magellanmeta-ai/magellan-exist-glossary]
I’m happy to walk through the architecture, trade-offs, or how it could be extended into a RAG or data-catalog use case.
About the author: Loren Cahlander is a data architect and knowledge-graph practitioner with experience across healthcare, federal, and enterprise technology organizations. He has contributed to W3C standards communities and open-source semantic-web projects.
Top comments (1)
The taxonomy angle is interesting because AI-readiness often gets framed as "add embeddings" when the harder work is making the knowledge model explicit. SKOS-XL already gives you a vocabulary discipline that a model can use instead of guessing from loose strings.
I would be especially careful with labels, aliases, and deprecated concepts. Those are exactly the places where search feels fine for humans but becomes ambiguous for downstream agents.