Curriculum vitae

Liam Chalcroft
liamchalcroft@gmail.com · https://liamchalcroft.com

Machine learning researcher working on medical and spectral imaging: domain generalisation, physics-constrained synthetic data, and self-supervised pre-training for 3D imaging. Currently Founding Computer Vision Scientist at Prospectral.

liamchalcroft@gmail.com Google Scholar ORCID GitHub

Experience

Founding Computer Vision Scientist
Prospectral, London
- Lead machine learning research and its integration into the wider product.
- Build production systems spanning spectral sensing, computer vision and model deployment.
- Own technical strategy for spectral imaging AI.
2025–present
Computer Vision Researcher
Tractive, London
- Led ML research at an a16z-backed pre-seed startup applying 3D generative AI to retopology.
- Trained transformers at scale with PyTorch and FSDP on Google Cloud.
- Wrote production backend code in C++ and Rust.
2024
PhD Researcher
Wellcome Centre for Human Neuroimaging, University College London
- Built a physics-constrained synthetic data framework for stroke lesion segmentation that transfers to unseen clinical scanners and sequences.
- Designed convolutional attention architectures for 3D segmentation, presented at the Medical Imaging meets NeurIPS 2023 workshop.
- Developed sequence-invariant contrastive pre-training for 3D MRI encoders.
- Contributed to the ISLES'22 challenge ensemble published in Nature Communications.
2021–2026
MRes Researcher
University College London
- Built hypernetwork-based segmentation conditioned on imaging domain.
- Studied image-level false positives in segmentation, published at the ASMUS workshop at MICCAI 2021.
2020–2021
Research Scientist, Intern
Schlumberger Cambridge Research, Cambridge
- Characterised non-Newtonian drilling fluids by rheology and diffusing-wave spectroscopy.
2018–2019

Education

PhD, Machine Learning
University College London
- Thesis: Robust Deep Learning for Stroke Detection in Clinical Neuroimaging.
- Supervised by Prof. John Ashburner and Prof. Cathy J. Price FRS.
- i4health CDT (EPSRC and Wellcome).
2021–2026
MRes, Medical Imaging
University College London, Distinction
2020–2021
MSci, Chemical Physics
University of Bristol, First Class Honours
2016–2020

Teaching and supervision

Fellowship Project Supervisor
Fatima Fellowship
2023–2024
MSc Project and Research Supervisor
University College London
2022–2026
Tutor, Machine Learning and Data Science
Cambridge Spark
2022–2024
Outreach Project Supervisor
In2Research and University College London
2022
Teaching Assistant and Guest Lecturer
COMP0090 Introduction to Deep Learning, University College London
2021

Talks

Gradient-manifold alignment scheduling for physics-guided diffusion
SPIE Photonics Europe 2026, oral.
April 2026
Domain-Robust Deep Learning for Stroke Detection in Clinical Neuroimaging
yESMInar #37: AI in Medical Imaging, European Society for Molecular Imaging, invited talk.
April 2026

Publications

Tokenizer-Generator Coupling in Medical Image Generation. NeurIPS 2026.
2026
Gradient-manifold alignment scheduling for physics-guided diffusion. SPIE Photonics Europe 2026.
2026
GAZE: Grounded Agentic Zero-shot Evaluation with Viewer-Level Tools and Literature Retrieval on Rare Brain MRI. AIiH 2026.
2026
Robust Deep Learning for Stroke Detection in Clinical Neuroimaging. UCL PhD 2026.
2026
Self-directed multimodal AI: multi-turn reasoning vision–language models for rare brain pathology. RCR Open 2026.
2026
Unified 3D MRI Representations via Sequence-Invariant Contrastive Learning. SASHIMI 2025.
2025
Domain-Agnostic Stroke Lesion Segmentation Using Physics-Constrained Synthetic Data. MICCAI 2025.
2025
Synthetic Data for Robust Stroke Segmentation. MELBA 2025.
2025
DeepISLES: a clinically validated ischemic stroke segmentation model from the ISLES'22 challenge. Nature Comms 2025.
2025
LKA: Large-kernel Attention for Efficient and Robust Brain Lesion Segmentation. NeurIPS MedImg 2023.
2023
Development and evaluation of intraoperative ultrasound segmentation with negative image frames and multiple observer labels. ASMUS 2021.
2021

The full list, with abstracts and BibTeX

Skills

Python, Rust, C++, MATLAB.
languages
PyTorch, MONAI, SPM.
frameworks
Medical image analysis, domain generalisation, synthetic data, self-supervised learning, generative modelling, spectral imaging.
domains