James Oldfield

James Oldfield is a final-year PhD student at Queen Mary University of London (QMUL), researching AI safety and interpretability. He currently holds roles as a Visiting Student at the University of Oxford and Research Associate at AIGI. His work focuses on decomposing machine learning models for human-interpretable insights and aligning AI outcomes with human values. He has been supervised by prominent researchers including Ioannis Patras and Mihalis A. Nicolaou.

Key affiliations include:

• QMUL: PhD in Electronic Engineering and Computer Science
• University of Oxford: Visiting Student (2025–)
• AIGI Oxford: Research Associate (2025–)
• University of Wisconsin–Madison: Honorary Associate (2024)

Research interests span AI interpretability, GANs, tensor methods, and scalable model specialization. Notable achievements include:

• Top Reviewer at NeurIPS 2024
• Publications at venues like NeurIPS, ICLR, and TPAMI
• Leading projects like PandA (parts/appearance decomposition in GANs)

He has contributed to teaching at QMUL, including roles as a Teaching Assistant for Deep Learning and Computer Vision courses.