research
Research background
Artificial intelligence (AI) has advanced in recent years due to innovations in machine learning and deep learning algorithms, computing hardware and software, and data availability. My academic research has a significant focus on machine learning for health, biology, and personalised medicine: these domains offer immense transformative potential for AI applications, yet even with larger datasets, the inherent characteristics of the data and complex needs of computational biology, bioinformatics and digital health applications require innovative approaches to achieve breakthrough results. Collaborating with international teams and utilizing large-scale, world-leading datasets such as the UK Biobank, Finnish national health registry data, and the FinnGen project, my work develops new probabilistic ML and deep learning strategies aimed at creating robust models that can effectively handle the complexities of real-world data and applications.
My recent projects include:
Synthetic Data for Personalised Medicine Research: Introduced an ML framework for generating synthetic but realistic datasets for genotypes and phenotypes, collaborating with the Europe-wide INTERVENE consortium to enable researchers to test new computational methods while protecting sensitive health information.
Geometric Deep Learning for Family Networks: Developed a novel deep learning approach using graph representation learning to predict health outcomes in families, working with the Institute of Molecular Medicine Finland on nationwide electronic health records (EHRs) and family networks for over 7 million patients.
Meta-learning for Health Record Tasks: Studied meta-learning approaches that "learn how to learn" from related machine learning tasks and how causal task similarity affects the generalizability and negative transfer properties of these algorithms for health-related prediction tasks. This work included a case study for stroke prediction using the UK Biobank and FinnGen datasets.