William Thorne
Researcher Phd Candidate
United Kingdom
Summary
👤
Senior🎓
Top SchoolWilliam Thorne is a PhD candidate in natural language processing at the University of Sheffield working with the National Gallery to structure and enhance search across art-historical text records. He specialises in low-resource language modeling—quantisation, low-rank adaptation, PEFT and retrieval-augmented approaches—aimed at practical GLAM deployments where compute, memory and labelled data are constrained. His current research pioneers unsupervised graph-to-text-to-graph translation via iterative back-translation to bridge natural language and Linked Open Data (Linked Art), reducing manual curation overhead and making LOD more accessible. Complementing his academic work, he develops language-model microservices for the self-hosted AI assistant Lovey and consults on in-game AI at Parable Studios, shaping models with creatives in mind rather than replacing them. With a decade of technical experience and hands-on work across research, engineering and applied ML, he blends deep academic methods with production-aware implementation.
10 years of coding experience
Doctor of Philosophy - PhD, Natural Language Processing, Doctor of Philosophy - PhD, Natural Language Processing at The University of SheffieldA-Level, Computer Science, A, A-Level, Computer Science, A at Ralph Allen School
Stackoverflow
345reputation
123kreached
8answers
4questions
Github Skills (58)
data-science9
lifecycle9
artificial-intelligence9
lightning9
pytorch9
scaling9
python9
machine-learning9
language-modeling9
ai9
deep-learning9
nlp8
esp328
transformer-architecture8
arduino8
Programming languages (4)
Github contributions (5)
wrmthorne/cycleformers
Jul 2023 - Jan 2025
A Python library for efficient and flexible cycle-consistency training of transformer models via iteratie back-translation. Memory and compute efficient techniques such as PEFT adapter switching allow for 7.5x larger models to be trained on the same hardware.
Contributions:17 PRs, 87 pushes, 20 branches in 1 year 6 months
cycle-consistent-learninghuggingface-transformersiterative-back-translationpeft-fine-tuning-llmpytorch
wrmthorne/AQG-From-Human-Feedback
Oct 2023 - Dec 2023
Contributions:219 pushes, 1 branch in 2 months
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.