Semantic Architect

Dyad is seeking a Semantic Architect to design and operationalise our graph-integrated generative AI architecture. This is a senior, hands-on technical leadership role within the Applied AI function. The Semantic Architect is responsible for ensuring that unstructured clinical text, structured knowledge (ontologies and graphs), and generative AI systems work together as a coherent, production-ready system. You will bridge NLP pipelines, LLM-based reasoning, and knowledge graph grounding to produce outputs that are accurate, explainable, and suitable for use in regulated healthcare environments. The role combines architectural ownership with day-to-day technical coordination and is critical to scaling Applied AI delivery without creating single points of technical dependency. This position is offered on a hybrid basis from our London office.

Core responsibilities

• Technical leadership & architecture ownership
  • Design and own end-to-end AI architectures that integrate: LLM-based reasoning and orchestration, pipeline evaluations and benchmarking, knowledge graph grounding and validation.
  • Define how structured semantics constrain, validate, and guide generative outputs.
  • Make pragmatic architectural decisions balancing accuracy, performance, explainability, and engineering effort.
  • Set standards for system design patterns across the Applied AI stack.
  • Ensure AI features are production-ready, robust, and aligned with product intent.
• Day-to-day technical coordination
  • Coordinate technical work within the Applied AI team.
  • Break product requirements into coherent, technically sound implementation plans.
  • Ensure alignment between NLP components, graph systems, and application layers.
  • Maintain architectural coherence as features evolve and scale.
  • Represent Applied AI in cross-functional technical discussions with Engineering and Product.
  • Define and maintain evaluation frameworks for precision and recall of extracted clinical concepts and regression testing across model updates.
  • Implement structured output approaches (e.g. schema-constrained generation, ontology-driven formats).
  • Design iterative feedback loops, including human-in-the-loop review where appropriate.
  • Ensure measurable improvements in grounding, explainability, and reliability over time.
• Compliance-aware AI engineering
  • Design AI workflows that embed traceability, auditability, and data minimisation by default.
  • Ensure architectural decisions align with medical device and data protection requirements across UK and US contexts.
  • Work proactively with Clinical Safety and QARA teams to avoid late-stage architectural risk.

Requirements

• A minimum of a master's degree in computer science with an AI focus or equivalent is required, as well as at least 3+ years commercial experience delivering knowledge-based systems in real production environments.

Core technical expertise

• Strong hands-on experience in designing production AI systems that integrate LLMs with structured knowledge.
• Deep understanding of trade-offs between symbolic reasoning, probabilistic inference, and generative pattern matching.
• Experience building systems that combine NLP pipelines with structured data validation or knowledge graphs.
• Strong Python experience for NLP pipelines, LLM orchestration, evaluation tooling, and data processing.
• Experience integrating AI systems into production services (Elixir experience is a plus, or willingness to engage deeply with it).
• Experience with prompt engineering using structured outputs.
• Familiarity with schema-constrained generation (e.g. JSON or ontology-driven outputs).
• Experience designing evaluation and benchmarking frameworks for production LLM systems.
• Understanding of model versioning, regression testing, and iterative improvement cycles.

Knowledge graph integration

• Experience designing AI pipelines that are constrained or validated by graph structures.
• Ability to collaborate effectively with Knowledge Engineers to ensure graph representations are AI-usable.
• Understanding of performance and scaling considerations when integrating graph-backed validation.

Operating context

• Experience working in regulated or high-assurance environments is strongly preferred.
• Ability to balance experimentation with production discipline.
• Comfortable operating in a fast-moving startup environment with high ownership expectations.

Personal attributes

• Systems-oriented thinker who values coherence over novelty.
• Pragmatic builder rather than research-focused experimentalist.
• Comfortable taking technical ownership and accountability.
• Strong communicator who documents and disseminates architectural knowledge to avoid bottlenecks.

Benefits

• Company pension
• 25 days of paid annual leave (pro-rata)
• Flexible hybrid working environment
• Employee Assistance Programme
• Modern, dog-friendly office near Chancery Lane with free drinks