Research
We study how intelligent systems reason, decide and operate when the evidence is complex, the rules matter and errors have consequences.
Our programme pages distinguish active research, applied work and proposed outputs. Publication status is never implied.
Domain intelligence for financial language, structured data and institution-specific knowledge.
Central question
How can a model reason across policy, transaction and customer context without losing provenance?
Methods
Expected outputs
Agents that can assist, analyse and act within explicit financial and operational boundaries.
Central question
What must an agent know, verify and record before it can safely participate in a regulated workflow?
Methods
Expected outputs
Decision systems that connect predictive performance with explainability, fairness and portfolio outcomes.
Central question
Can adaptive credit systems improve access and timing while keeping risk decisions reviewable?
Methods
Expected outputs
Evaluation and governance methods for systems deployed where errors carry financial consequences.
Central question
How should reliability, traceability and human accountability be measured before deployment?
Methods
Expected outputs
Method
Define the decision, affected people, institutional constraints and acceptable uncertainty.
Develop the smallest model or system that can test the research question honestly.
Test behaviour across normal, ambiguous, adversarial and distribution-shift scenarios.
Introduce controls, human review and monitoring proportional to consequence.
Study real outcomes and feed the evidence back into models, interfaces and policy.
Research archive