Hi! I’m Paul. I am a PhD student in the Language Technology Lab at the University of Cambridge, focusing on Modular and Efficient Deep Learning for NLP under supervision by Ivan Vulić and Edoardo Ponti.

I previously obtained my Master of Informatics (MInf) in 2024 from the University of Edinburgh, where I worked on Distributed Training and Cross-Architecture Knowledge Distillation under supervision by Hao Tang.

When I’m not studying or researching deep learning methods, I am probably on a run🏃, taking photos📷 or travelling somewhere⛰️.

Research Interests

• Modular Deep Learning (e.g. merging models without retraining)
• Parameter-efficiency in training, fine-tuning and inference (e.g. adapters, MoEs)
• Training paradigms for efficient and distributed training
• Computational Neuroscience: Biologically grounded models of cognition, behaviour and illnesses

Previous work

1. Decoupling the Method of Auxiliary Coordinates to Distributively Train Deep Neural Networks.
   Master’s Thesis, supervised by Hao Tang.
   View PDF →

2. Cross-Architecture Knowledge Distillation Techniques for Models with Mismatched Output Dimensions.
   Bachelor’s Thesis, supervised by Hao Tang.
   View PDF →

3. Review of a Biologically Inspired Neural Network to Model PTSD and Eye Movement Desensitisation Reprocessing Therapy.
   Coursework in Computational Cognitive Neuroscience, taught by Peggy Seriès.
   View PDF →

4. Bioinformatics Analysis of Autism-Related Gene Networks.
   Coursework in Bioinformatics, taught by Ian Murray.
   View PDF →

Other experience

1. Teaching Assistant for Machine Learning @2023/24

2. Research Assistant intern for Kartic Subr, working on Graph Neural Networks for spectral coarsening of 3D meshes. @Summer 2023

3. Tutor for Machine Learning @2022/23

4. Exchange year at the University of Hong Kong @2021/2022

---