ASCEND research arc
Modelling clinical trajectories to study risk, patterns of disease, and how clinical models can be interpreted.
Latest public milestone ·
Interpreting clinical sequence models (preprint)
Clinical AI · EHR representation · research software
Chris Sainsbury
Clinician-researcher working on clinical foundation models, EHR representation learning, translational AI, and research software.
I work across medicine, machine learning, and research infrastructure. This site is a public front page for the main things I am working on: active projects, recent outputs, papers, software, and current directions.
Selected work
Featured projects
The questions behind the work, with milestones linked to public outputs.
NEXUS
Investigating where ideas and methods from one field might help solve problems in another.
Public release
glucose.ai
A public literature-discovery site connecting diabetes research with developments in AI and machine learning.
pi / Cog workflows
Tools for connecting research sources, experiments and synthesis, so useful work is easier to organise and revisit.
Papers · software · public outputs
Public milestones
Selected developments, using the dates recorded by their linked sources. Preprints are work in progress, not peer-reviewed findings.
· Preprint
Interpreting clinical sequence models
A preprint on sparse autoencoder analysis of clinical sequence model representations is available on arXiv.
· Preprint
FlatASCEND preprint
Submitted on 13 April 2026, this preprint studies clinical sequence generation and observational pharmacological associations. It is available on arXiv.
· Preprint
ASCENDgpt preprint
A preprint on phenotype-aware cardiovascular risk prediction from electronic health records is available on arXiv.
recent outputs
Papers and scholarly outputs
Causal counterfactual simulation for treatment decisions in multimodal lung disease data
Y Zhu, L Zhang, C Sainsbury, F Dong, J MacLay, DJ Lowe, X Ye
Computers in Biology and Medicine 213, 111807
Exploring the use of AI-generated counterfactual chest X-rays to enhance diagnostic learning in medical education
G Mohr, Y Zhu, X Ye, M Lennon, C MacLellan, J Maclay, DJ Lowe, ...
BMC Medical Education
Sparse Autoencoder Decomposition of Clinical Sequence Model Representations: Feature Complexity, Task Specialisation, and Mortality Prediction
C Sainsbury, F Dong, A Karwath
arXiv preprint arXiv:2605.04072
areas of work
What connects the portfolio
The portfolio spans academic papers, translational grants, scientific-discovery systems, and software for organising research. The aim is not just to build models, but to make them clinically meaningful and operationally usable.