Semantic Architect

Overview

The Semantic Knowledge Architect joins an established delivery team that includes an AI engineer responsible for agent development. The two roles work in close partnership: the AI engineer builds and maintains the agents; this role owns the knowledge layer those agents depend on.

Your responsibility is the quality, integrity and evolution of the semantic knowledge layer: the ontologies, taxonomies and knowledge graph structures that determine what agents know, how they connect information, and whether their outputs are accurate. You will expand and govern structures that are already in place, working iteratively as the programme develops and agent use cases grow. You will also be the diagnostic layer between agent outputs and the knowledge layer: when an agent produces an incorrect or incomplete output, you identify whether the root cause sits in the knowledge structure and fix it s is a technically precise, high‑accountability role.

The accuracy of agent outputs across the group depends directly on the quality of the knowledge layer you maintain.

The role follows a continuous iterative cycle across four activities:

• Expand: take new or evolving business and technical requirements and extend the existing ontology and taxonomy to accommodate them, maintaining consistency with the established model
• Govern: manage the ontology and taxonomy as controlled, versioned assets with documented change rationale, ownership and review cycles
• Validate: review agent outputs in collaboration with the AI engineer, identify gaps or inaccuracies that originate in the knowledge layer, and trace them to their structural source
• Iterate: update ontologies, taxonomies and graph structures based on validation findings, closing the loop between agent performance and knowledge quality

You will work closely with other team members for data context and domain knowledge. The data landscape is already understood within the team, so onboarding to the knowledge environment will be well supported.

• Extend the existing ontology to reflect new business requirements, additional entities and evolving operational concepts
• Expand and maintain the enterprise taxonomy, ensuring classification remains accurate, consistent and fit for agent consumption
• Own the governance framework for both the ontology and taxonomy: versioning, change control, documentation and review cadence
• Work with the AI engineer to review agent outputs and identify where knowledge‑layer gaps or inconsistencies are driving errors
• Update ontological and taxonomic structures in response to validated agent performance issues
• Maintain the knowledge graph as an accurate, traversable semantic layer connecting group operational data
• Ensure data ingestion into systems is governed by clear metadata and semantic standards
• Document all structural decisions, changes and rationale to support long‑term knowledge asset governance
• Contribute to the broader delivery team, sharing knowledge context with data, platform and business colleagues as needed

• Ontology Expansion & Maintenance:
  Build upon and extend the existing ontology foundation. This is not a greenfield task. The core domain model exists and the data is mapped. Your role is to deepen, refine and evolve it as operational requirements develop.
• Graph data modelling
• Entity modelling and resolution
• Semantic layer design
• Graph databases:
  Neo4j, Stardog, Graph
  
  DB, Amazon Neptune
• Data integration and linkage

• Extended knowledge graph covering expanding operational domains
• Semantic integration models connecting data sources accurately
• Entity relationship frameworks that agents can reliably traverse
  
  Graph integrity standards and validation processes

• Ontology engineering and extension
• Concept and semantic modelling
• RDF/OWL/SKOS
• Reasoning frameworks
• Version control for ontology assets
• Conflict resolution within existing models

• Extended domain ontologies aligned to evolving business requirements
• Refined concept models with documented change rationale
• Versioned semantic schemas with change history
• U…