Agentic AI & human–AI teaming
LLM agents, RAG, tool use, LangGraph orchestration, evaluation, approval gates and safe automation. Research includes human reliance and experimental design for operational settings.
HITLRAGLangGraph
Research direction
Six intersecting research themes
From responsible foundation models to circular systems and biomedical analytics.
Responsible AI & societal adoption
Fairness, accountability, explainability, creative labour, consumer discourse, bias, privacy, trustworthy deployment and evidence-led public-sector governance.
AI ethicsFairnessEvaluation
AI, IoT & circular economy
Smart-product tracking, knowledge-driven circular decision support, intelligent waste classification, digital twins and resource-efficient business models.
Industry 4.0/5.0IoTDigital twins
Computer vision & applied ML
Explainable image classification, transfer learning, hybrid data, feature engineering and model evaluation across sustainability and biomedical microscopy.
PyTorchCNNsXAI
Trade, infrastructure & decisions
Data-driven supply chains, physical/digital assurance, interoperable trade documents and decision support connecting real-world sensing to trustworthy data.
Digital tradeRFIDData quality
Data science for business & services
Statistical modelling, predictive analytics, NLP, operational insights, evidence-based recruitment strategy and organisational AI readiness.
StatisticsNLPBusiness AI
2026 Research spotlight
Separating the art from the algorithm.
Co-authored research published in the Proceedings of Machine Learning Research analyses the tension between the perceived usefulness of generative AI and concerns about how models are trained. The work uses human annotation and NLP to examine 7,371 Reddit posts across professional communities.
Read the peer-reviewed paperWorking methods
From question to robust evaluation
01
Frame the problem
Understand users, stakes, data, stakeholders and constraints.
02
Build appropriately
Choose statistical, ML, deep learning or agentic approaches on evidence.
03
Validate openly
Evaluate quality, robustness, bias, failure modes and meaningful outcomes.
04
Keep people in control
Document governance, monitoring, accountability and feedback.
Collaborate / Learn / Build
Make research useful in the real world.
Open to research collaboration, speaking, executive training, responsible AI advisory and industry-facing projects.