King's College London · Department of Biostatistics and Health Informatics
A new lab at King's College London, opening Q3 2027.
The DREAMS Lab will be established in the Department of Biostatistics and Health Informatics at King's College London, funded by a Wellcome Career Development Award. Recruitment begins in Q3 2027, for one PhD studentship and, over the first five years, three postdoctoral positions.
The data
Sixteen hospitals, four countries
The lab works from records hospitals collect in routine care rather than from curated research cohorts. Models are trained across sites by federated learning, so the data never leave the institution that holds them.
- Hospital centres
- 16Twelve in the UK, plus the Netherlands, Canada and South Korea.
- People with MS
- 37,650Routine clinical records, not a curated research cohort.
- MRI studies
- 169,425112,950 brain studies and 56,475 spinal cord acquisitions.
- Clinical notes
- 376,500Letters written by different clinicians, in four languages.
Role 1 · Q3 2027
PhD studentship
Doctoral research on the curation and harmonisation of multi-centre MRI and clinical records, held in the Department of Biostatistics and Health Informatics at King's College London.
The records held across the sixteen centres were not collected for research. The studentship concerns the methods by which they can be linked, harmonised and pseudonymised into a resource that models can be trained and evaluated on, and by which that resource can be documented and quality-controlled at scale. Those methods form the substance of the thesis, and the appointee is expected to complete a PhD alongside the research.
Responsibilities
- Build the harmonised, pseudonymised dataset on which the lab's other work packages depend, across sixteen centres in four countries.
- Establish and validate the de-identification pathway for images and for free text.
- Develop a first-author line of work through to publication as the doctoral component.
Essential
- A degree in computer science, engineering, physics, mathematics, statistics, neuroscience or medicine, or equivalent experience.
- Python, with version control, testing and reproducible environments.
- Working knowledge of a deep-learning framework, PyTorch preferred.
- Eligibility to work in the United Kingdom and to register for a PhD at King's College London.
Desirable
- Experience of clinical or hospital data and the information governance that accompanies it.
- Medical image formats and MRI preprocessing pipelines.
- De-identification, including DICOM anonymisation and free-text de-identification.
- Natural language processing applied to clinical text.
Role 2 · from Q3 2027
Postdoctoral position
Three positions over the first five years, spanning large-scale data curation, neuroimaging, clinical modelling and multimodal algorithm design.
Clinical machine learning is usually developed on curated datasets and performs poorly on routine hospital data. The lab begins from the hospital data instead and treats robustness as a research question. The work packages were written to have largely independent outputs, so that a postdoctoral researcher contributes to the collaborative programme while developing an independent line of research.
Posts are based at King's College London, with clinical work alongside King's College Hospital and part of the MRI modelling with the Montreal Neurological Institute. Visiting contracts at collaborating hospitals are to be expected, since federated learning takes the computation to the data.
Essential
- A PhD, awarded or submitted, in machine learning, computer vision, natural language processing, medical image analysis, biostatistics or a related field.
- Strong Python and PyTorch, including multi-GPU training.
- A record of peer-reviewed publication, and publicly available code.
- Eligibility to work in the United Kingdom.
Depth in one or more of the following
- Self-supervised representation learning applied to 3D volumes.
- 3D medical image segmentation and longitudinal imaging biomarkers.
- Encoder pretraining and concept-level representation for clinical text.
- Vision–language alignment, or federated learning across hospital sites.
What the post offers
- An independent work package with its own outputs, alongside collaborative work across the team.
- Access to routinely collected clinical data from sixteen hospitals in four countries.
- Collaboration with the Montreal Neurological Institute, and with clinical neurology at King's College Hospital.
How to register interest
Send a CV to arman.eshaghi@kcl.ac.uk
The posts are not yet open for applications. They will be advertised formally through King's College London, with salary, grade and contract terms, when the lab opens in Q3 2027.
Expressions of interest sent before then are held on file, and those who write will be contacted before the adverts are published. Please send a CV and a short covering letter stating whether you are writing about the studentship or a postdoctoral position, with links to any code or publications. Enquiries from people of backgrounds under-represented in machine learning and in academic research are particularly welcome.