Biography
With an MMath in Mathematics and Statistics (2015) and a DPhil in Clinical Medicine (2020) funded by an NDM Prize Studentship, University of Oxford, Karina is a truly multidisciplinary researcher. Her doctoral research comprised of using electronic health records to improve management of E. coli bloodstream infections, using traditional epidemiological methods; and predicting antimicrobial resistance in such infections by using machine learning methods and bacterial genome-wide association studies.
As a statistician and machine learning researcher she has also worked on the Office for National Statistics Coronavirus Infection Survey, on different types of surveillance models; and on understanding reported symptoms over time, and which combinations of symptoms are most predictive of Sars-CoV-2 positivity. She has had a Mathematical Lectureship with Trinity College and Corpus Christi College, teaching the Probability and Statistics courses. As part of the Computational Health Informatics lab she will be leading the NIHR Healthcare Protection Research Unit programme.
Research Interests
Karina is interested in using state of the art machine learning techniques on routinely collected electronic health records and bacterial whole genome sequencing data in order to improve the management of infectious diseases. Her main drive is solving real clinical problems, but with a background in Mathematics and Statistics she is also very keen to further develop methods if the data require it.
Current Projects
She is keen on exploring four big areas: how can routine surveillance of healthcare-associated infections be automated optimally, can we predict antimicrobial resistance based on what is being used today and where antimicrobial resistance is, can we predict multivariate genetic mechanisms from phenotypes and individualising antibiotic treatment. In order to tackle these challenging questions she is going to hone in on the strengths of deep algorithms for time-series, meta-learning and multi- task classification, as well as explore time-series clustering algorithms for unsupervised learning settings.
