PhD Researcher - AI TokCart

Project background

The healthcare systems of the future must harness data effectively to support clinicians, allowing them to focus on patient care while leveraging AI to detect patterns beyond human perception, enhance diagnostic accuracy, optimise workflows, improve risk assessment and communication. Developing AI models that address these needs is particularly urgent in ageing societies, where rising patient numbers coincide with increasing workforce constraints.

A major contributor to disability in older adults is impaired musculoskeletal (MSK) health, with fall-related fractures representing a central research focus for our group. Despite their substantial socioeconomic burden, progress in prevention and screening has been limited, partly because fall and fracture risks are often addressed independently, despite their shared underlying mechanisms. A promising direction for AI in healthcare is the seamless integration of clinical image-based foundation models that can be fine-tuned across multiple use cases.

In this project, we aim to develop cross-modal AI models for MSK health evaluation by linking diverse clinical imaging datasets. The central innovation is the creation of a shared, tokenized latent space that enables data translation, risk modelling, and local deployment, even in settings with limited imaging data. Particular emphasis will be placed on integrating advanced biomechanical markers of bone health into AI workflows, supporting more precise and clinically actionable risk assessment.

Job description

The PhD candidate will be embedded into a team of researchers that is working on developing large medical image-based AI models for managing MSK health.

Some key components in the work are:

• Pretraining, adapting, and evaluating foundation models for medical image data using self-supervised and multimodal representation learning.
• Developing cross-modal image modelling methods, including cross-modal translation, multimodal fusion, and related generative or representation-learning approaches.
• Evaluating the learned models on clinically relevant downstream tasks, such as bone health assessment, anatomical and biomechanical biomarker estimation, and fracture-risk-related prediction.
• Co-supervision of MSc and BSc students.

Please note that the employment will be at the Singapore-ETH Centre in Singapore, and local working regulations will apply.

Funding is provided for up to four years. After successful completion of studies, the PhD degree will be awarded by ETH Zurich.

Profile

• Applicants should hold an MSc degree and have a strong background in computer science, engineering, mathematics, physics, or a related quantitative discipline.
• Candidates should be capable of working largely independently on complex research topics and demonstrate strong motivation.
• We are particularly interested in applicants with a solid foundation in machine learning and AI, image processing, strong Python programming skills, and experience in software development.
• Background knowledge in parallel computing or high-performance computing is desirable.
• Experience with software engineering practices, including version control (e.g., git) and collaborative development workflows, is required.
• Applicants must be proficient in written and spoken English.