• Published a paper in the Pattern Recognition and Artificial Intelligence (PRAI) conference titled "Raw Instinct: Trust Your Classifiers and Skip the Conversion" • Conducted research and analysis of soil moisture datasets to explore their potential use for ReDoCO2 project, aimed at mapping peatlands in detail and accurately estimating CO2 emissions and carbon stocks• Utilized Bayesian Deep Learning to perform uncertainty estimation in Graph Neural Networks, ensuring robust and reliable predictive modelling.

Title: Speech Enhancement On An External Device• Conducted research on state-of-the-art real-time speech enhancement methods.• Focused on leveraging abundant resources in mobile devices for enhanced performance compared to limited resources in hearing aids.• Trained and converted relevant models, further utilizing them in an Android application.• Conducted comprehensive comparative testing between PC and mobile models.