I am a postdoctoral researcher at the HCC group, Department of Computer Science, University of Oxford, working with Prof. Sir Nigel Shadbolt and Prof. Sir Tim Berners-Lee (at the EWADA project), on user-centered data usage control and user-centered AI, on decentralized Web or alike contexts, especially SoLiD (Social Linked Data).
Prior to this, I obtained PhD degree from AIAI, School of Informatics, the University of Edinburgh, supervised by Prof. Malcolm Atkinson, Dr. Petros Papapanagiotou and Prof. Jacques Fleuriot.

In general, my research surrounds data governance, data usage control and data protection mechanisms, especially in decentralized / multi-stakeholder contexts. It mainly concerns the technical mechanism design, but also considers user perspectives and expectations.
(Neural-)Symbolic AI, Knowledge Graph and Privacy-Enhancing Technology are the main topics for the technical parts. Data, accountability and user (empowerment) are the main factors.
More specifically, the following themes are covered:

  1. Data Terms of Use (DToU): How do we make data usage more accountable and controllable (and easy), for all stakeholders, especially regular people? [ref]
    • Regulatory / Compliance: what is the status quo, and (how) can we improve that?
  2. Data privacy: How to securely and efficiently use multiple users’ data (collaboratively) in a decentralised data architecture? [ref]
  3. GenAI: What challenges and opportunities do Large-Language Models (LLMs) or personal agents bring (to facilitate people)? [ref]
  4. User perception: how do users perceive (the issues of) data and how can we facilitate them?

Feel interested? Find out more from my papers in the publication page.

Linking back to PhD work

The DToU topic also bridges back with my PhD work on modelling and reasoning of data-use policies, in research / collaboration contexts (e.g. e-Science, scientific workflows), with the following questions:

  1. How to formally model data-use policies taking the properties of decentralised contexts into account?
  2. How to deal with the Multi-Input-Multi-Output nature of the workflows in such data-processing contexts?
  3. How to derive the data-use policies for the output data?
  4. How to check compliance of such policies?

See my previous research homepage for more details if interested.