uring my bachelor's studies, I wrote a thesis on time-series prediction using Support Vector Machines (SVMs). Working on this topic sparked a deep interest in time-series analysis, which became the foundation for my two master's theses. One focused on the applications of neural networks and deep learning in financial time-series processing, while the other explored the use of Hidden Markov Models (HMMs) for analyzing these time-series.

Through these projects, I gained valuable insights into how the choice of a model depends on the specific goals of a task. Deep learning models achieved higher accuracy but required significant computational resources and training time. In contrast, HMMs, while less accurate, offered faster training times and greater interpretability.

This experience ignited my fascination with Bayesian modeling. Hidden Markov Models, in particular, intrigued me due to their discrete latent space and generative nature. Unlike discriminative models, HMMs not only enable classification but also facilitate efficient sampling of synthetic data and provide insights into the hidden structure of the data, such as state-switching probabilities and emission properties.

Outside of academia, I have worked in the industry with companies like Microsoft, Opera Software, and NavAlgo. These roles allowed me to hone my coding skills and gain practical experience in deploying machine learning models in real-world scenarios. This combination of academic research and industry experience has shaped my expertise and deepened my passion for machine learning and time-series analysis.