Were you interested in health data science before coming to the UK?
I completed my undergraduate degree in bioinformatics in Spain, which was quite a new field at the time. I鈥檝e always been fascinated by the intersection of health and biomedical science with mathematics, computer science, and technology. Bioinformatics gave me the perfect foundation to explore the kind of interdisciplinary science that underpins health data research.

During my bachelor鈥檚, I began working at the BCN-AIM Lab, a research group specialising in the development of trustworthy AI systems for healthcare. I continued there alongside my first master鈥檚 degree in computational biomedical engineering. Over those four years, I contributed to several European Commission, funded Horizon projects such as EarlyCause, LongITools and STAGE, developing predictive models for a variety of diseases using large-scale exposome datasets.

This work was not purely technical. I was also actively involved in developing ethical frameworks and policy recommendations for AI in healthcare, including contributions to the FUTURE-AI international consensus guidelines and a WHO report on risk communication of ambient air pollution. That experience shaped my conviction that AI research must be closely tied to policymaking, so that innovations translate into fair, effective, and trusted real-world health systems.

Tell us about your master鈥檚 degree.
I pursued an MPhil in Population Health Sciences at the University of Cambridge because I wanted to deepen my expertise while still engaging with taught courses but also to be fully immersed in research. The programme offered the perfect balance, combining advanced training in epidemiology, statistics, and health data science with the opportunity to carry out my dissertation in a leading research lab.

I carried out my project at the Victor Phillip Dahdaleh Heart & Lung Research Institute, under the supervision of Dr. Samuel Lambert. My dissertation, Trustworthy Machine Learning Prediction of Long-Term Conditions in Diabetic Patients: Integrating Exposome and Genome Data from the UK Biobank, focused on building predictive models for multi-morbidity in individuals with type 2 diabetes. The research integrated multi-modal data, including clinical records, environmental exposures, and genomic information.

A central methodological component was the use of classifier chains to capture dependencies between co-occurring conditions, enabling joint disease prediction and improving model realism for clinical settings. A key emphasis was on ensuring the models were fair, explainable, reusable, and environmentally responsible. This included applying fairness metrics to detect biases, using SHAP for interpretability, calibrating predictions for clinical reliability, and tracking the carbon footprint of model development.

The approach was benchmarked against existing cardiovascular disease risk tools such as QRisk3, demonstrating the potential for more comprehensive and equitable risk prediction. The findings contribute to the growing evidence base for integrating genomic and exposomic data in predictive modelling for chronic disease management. I am now expanding this project for submission to a peer-reviewed journal.

And how did the scholarship fit in with that?
It was ideal. I was particularly interested in multi-morbidity, the fact that many people don鈥檛 just develop one condition, but multiple chronic conditions. This is especially relevant for people with type 2 diabetes, as their health trajectories can be quite different. The scholarship allowed me to explore this in depth.

It also made studying here financially possible. As an international student, I wouldn鈥檛 have been able to afford it otherwise. It provided not only financial support but also a network of experts and potential collaborators. Having two funders, Health Data Research UK and Diabetes UK, also meant recognition from two respected organisations in my research area.

Do you want to stay in the same area of research?
Yes. I want to continue working in multi-morbidity predictive modelling with a strong ethical focus, ensuring AI is fair, explainable, reusable, and environmentally responsible. A key part of my approach is integrating multi-modal data, including genetic information and exposomics, to better understand disease trajectories and improve patient care. I鈥檓 particularly committed to making sure that, when these models are implemented in clinical practice, they are free from biases that could disadvantage certain groups and that they truly support equitable, personalised healthcare.

What do you want to do in the future?
I want to continue in research and, one day, lead my own lab. That鈥檚 my dream. My next step is starting this October a fully funded PhD in Public Health and Primary Care at the University of Cambridge.

I have been fortunate to receive multiple competitive national and international awards, including funding from AstraZeneca, Fundaci贸n Rafael del Pino, and my Cambridge college Emmanuel. This recognition has been invaluable in building my research profile and advancing my career in health data science.

Alongside my scientific goals, I am deeply committed to shaping health policy. My past work with WHO and EU projects has shown me how research can influence regulatory frameworks, public communication, and ethical standards for AI in medicine. Going forward, I want my research not only to advance predictive modelling for multi-morbidity, but also to directly inform clinical decision-support systems, national health strategies, and global policy discussions on the responsible deployment of AI in healthcare.

  • Find out more about Marina聽here.