About the Lab
We advance research and methods that make AI dependable in regulated settings. Our work sets practical standards for reliability and transparency so organisations can adopt AI with confidence.
About the Lab
We advance research and methods that make AI dependable in regulated settings. Our work sets practical standards for reliability and transparency so organisations can adopt AI with confidence.
About the Lab
We advance research and methods that make AI dependable in regulated settings. Our work sets practical standards for reliability and transparency so organisations can adopt AI with confidence.
About the Lab
We advance research and methods that make AI dependable in regulated settings. Our work sets practical standards for reliability and transparency so organisations can adopt AI with confidence.
About the Lab
We advance research and methods that make AI dependable in regulated settings. Our work sets practical standards for reliability and transparency so organisations can adopt AI with confidence.
Who we are
We bring together machine learning, linguistics and rule-aware (neurosymbolic) techniques to solve problems where “almost right” isn’t good enough. Our aim is to build AI you can check, trust and sign off.
Who we are
We bring together machine learning, linguistics and rule-aware (neurosymbolic) techniques to solve problems where “almost right” isn’t good enough. Our aim is to build AI you can check, trust and sign off.
What we do
We develop evaluation methods and auditability techniques for enterprise workflows. Our work concentrates on three outcomes.
What we do
We develop evaluation methods and auditability techniques for enterprise workflows. Our work concentrates on three outcomes.
Accuracy
Correct answers on well-defined tasks, with clear pass criteria for each task.
Accuracy
Correct answers on well-defined tasks, with clear pass criteria for each task.
Accuracy
Correct answers on well-defined tasks, with clear pass criteria for each task.
Consistency
The same answer for the same facts, even after updates, with targets that limit drift over time.
Consistency
The same answer for the same facts, even after updates, with targets that limit drift over time.
Consistency
The same answer for the same facts, even after updates, with targets that limit drift over time.
Explainability
Short reasoning with links to the exact source text or rule, built for quick review and audit.
Explainability
Short reasoning with links to the exact source text or rule, built for quick review and audit.
Explainability
Short reasoning with links to the exact source text or rule, built for quick review and audit.
Current focus
Active areas of work across high-stakes, document-heavy workflows.
Current focus
Active areas of work across high-stakes, document-heavy workflows.
Policy & interpretation
Extract clause meaning from complex documents and apply explicit rules deterministically.
Policy & interpretation
Extract clause meaning from complex documents and apply explicit rules deterministically.
Policy & interpretation
Extract clause meaning from complex documents and apply explicit rules deterministically.
Compliance & decisions
Validate against internal and external policies, then issue consistent approve, deny or escalate outcomes with rationale attached.
Compliance & decisions
Validate against internal and external policies, then issue consistent approve, deny or escalate outcomes with rationale attached.
Compliance & decisions
Validate against internal and external policies, then issue consistent approve, deny or escalate outcomes with rationale attached.
Disclosure & reporting
Ensure required information is complete, evidenced and audit-ready before submission.
Disclosure & reporting
Ensure required information is complete, evidenced and audit-ready before submission.
Disclosure & reporting
Ensure required information is complete, evidenced and audit-ready before submission.
Join us
We’re building a multidisciplinary team to make AI dependable in regulated industries. If you care about clear methods, measurable results and thoughtful engineering, we’d like to hear from you.
Join us
We’re building a multidisciplinary team to make AI dependable in regulated industries. If you care about clear methods, measurable results and thoughtful engineering, we’d like to hear from you.