We build the infrastructure around intelligent systems.
NormLogic grows out of Sireto’s broader work in AI engineering and regulatory knowledge infrastructure. The same team builds production AI systems with enterprises and public-sector organisations that need them to be understood, tested and trusted.
Capabilities.
Eight things we build well, on their own or as part of a larger system.
Automated regulatory compliance
Compliance checking systems that evaluate policies, controls and evidence against structured regulatory requirements, with the source provision attached to every finding.
Regulation as code
Legislation, standards and internal policies transformed into structured, machine-executable rules with provenance.
Agentic systems
Multi-agent architectures for research, extraction, validation and operational workflows.
Regulatory knowledge engineering
Regulatory ontologies, knowledge graphs, semantic models and domain knowledge representation.
LLM engineering
Structured generation, model routing, evaluation, prompt optimisation and model-independent architectures.
AI evaluation
Datasets, benchmarks and systematic testing for extraction, reasoning and domain-specific AI systems.
AI infrastructure
APIs, observability, authentication, orchestration and production infrastructure for AI applications, including our own gateway and contract platform, openapi.ai.
Secure enterprise integration
Connecting AI systems with existing identity, data and operational environments without turning critical business processes into opaque black boxes.
How we build.
The same principles shape NormLogic and the systems we build with clients.
Explicit over implicit
Knowledge that matters is written down as structure, not left inside model weights and prompts.
Provenance by default
Every assertion a system makes can be traced to the source that justifies it.
Model-independent
Architectures that route between models and survive the next model release.
Evaluated, not demonstrated
Datasets, benchmarks and regression tests decide whether something works. A convincing demo does not.
Open representations
Standards such as LegalRuleML and Akoma Ntoso over proprietary formats, so what we build stays interoperable.
No new black boxes
AI integrates with existing identity, data and operational systems without making critical processes opaque.
Typical problems.
Most engagements start with one of these. Many end up touching several.
Automated regulatory compliance
Evaluate policies, procedures and controls against the regulations that apply, with a traceable finding for every gap.
Regulation as code
Transform legislation and policies into structured, machine-executable rules linked to their sources.
Regulatory change management
Identify affected rules, obligations and systems when legislation or guidance changes.
AI governance infrastructure
Give AI systems and agents structured regulatory constraints linked to authoritative sources.
Knowledge graph engineering
Transform fragmented domain knowledge into connected computational models.
Agentic document processing
Use specialised AI agents to analyse complex document collections with validation and provenance.
AI architecture
Design robust architectures for organisations moving from AI experiments to production systems.
Moving from AI experiments to production?
Tell us what you are trying to build and what has to hold true for it to be trusted. We will tell you honestly whether we are the right team.