Ronan Docherty

About:

I'm a graduate from Imperial College London with a Master's degree in Physics who is currently doing a PhD in machine learning for materials science with the TLDR group at Imperial. I also work part-time at Polaron, a materials analysis & design startup. I enjoy puzzles, programming and making kitsch webapps.

Personal Projects:

Orcball (web): paper football with powerups, with P2P multiplayer. Written with Phaser + Typescript for the GameDev.js jam.

   Itch    Github

Avalog: an analog (pieces move continuously) chess game, featuring draft mode and online multiplayer. Written in Godot, available on Steam.

Steam    Itch    Github

Hyperchess: a webapp to create & play on non-Euclidean chessboards either OTB or via P2P connection. Written in Typescript.

Website    Github

Bucket: a webapp that allows the user to create and fill a 'paint-by-numbers' from an uploaded image. Vanilla JS/HTML/CSS, with GCP function for the conversion.

Website    Github

CSSlide: a slideshow editor using arbritrary transitions of HTML elements between different CSS states. Winner of the IC Hack 2020 best educational hack; written in Vue.js.

Website    Github    Video

Academic Projects:

ImageRep: uses structure of micrograph to give confidence bounds on measured volume fraction. Flask backend + React TSX frontend.

Website    Github

SAMBA: a trainable segmentation web-app for materials science phase segmentation. Uses Meta's SAM for labelling suggestions and a random forest backend for the segmentation.

Website    Github

pyramaterised: a Python library for creating and measuring Parameterised Quantum Circuits for near-term quantum computing applications.

PyPI    Github

ChiraSnap: a cheap, compact circularly polarised imaging system. Contains code and instructions for building the physical camera (including .stl files for the casing).

Github

Publications and preprints:

• A. Dahari, R. Docherty, S. Kench, S. J. Cooper, 'Prediction of microstructural representativity from a single image', arXiV preprint, 2024
• R. Docherty, A. Vamvakeros, S. J. Cooper, 'Upsampling DINOv2 features for unsupervised vision tasks and weakly supervised materials segmentation', NeurIPS, AI4Mat Workshop, 2024
• G. Lei†, R. Docherty†, S. J. Cooper, 'Materials science in the era of large language models: a perspective', RSC Digital Discovery, 2024
• R. Docherty, I. Squires, A. Vamvakeros, S. J. Cooper, 'SAMBA: A Trainable Segmentation Web-App with Smart Labelling', JOSS, 2024
• M. D. Ward†, R. Docherty†, L. Minion, X. Shi, K. Anson, G. Siligardi, J. Nelson, J. Wade, M. J. Fuchter, 'Development of low-cost, compact chiroptical imaging systems', RSC Nanoscale, 2024

Skills:

• Python
  • Libraries: numpy, scipy, scikit-learn, pandas, flask, etc.
  • Tooling: ruff, black, mypy
  • (Automated) CI/CD
• Machine learning: random forests, PyTorch, CNNs, vision transformers, (L)LMs
• Computer vision: scikit-image, kornia, feature extraction
• Game dev: Godot, Phaser.js
• Web dev: Javascript, Typescript, React (JSX, TSX), some VueJS
• Some lower-level experience, including C, FORTH, Assembly
• Physics and materials-science background, with some lab experience
• Master's student supervision; marking

Achievements:

• Polaron wins the Government's £1M AI prize
• Winner (Best educational hack) IC Hack 2020; Runner up at Oxford Hack 2019
• Best feature article, RSCU Science Challenge 2019
• Extracurricular university courses: pass with merit in French (CEFR-B1) and pass with distintion in creative writing
• National Citizenship Service Award