Andrei Prioteasa

This is a personal website to showcase some of my projects I’ve been involved in during my Bachelor’s and Master’s degree and to present my research interests, to the extent I am aware of them now. It’s a process.

My first contanct with Machine Learning and my subsequent academic journey was through Computer Vision. During an ERASMUS+ semester spent at TU Darmstadt, where I participated in professor’s Roth Computer Vision class, I started to think more seriously about Machine Learning. It was curios to me back then how this relates to the human vision apparatus and how it does not. I ended up doing my Bachelor Thesis on Explainability in the context of Human Action Recognition (which you can also find on my website). The why behind the how Machine Learning models take decision was starting to bug me and I wasn’t satisfied with the Black Box dogma. Since then, my interests have shifted, my motivation did not.

I’m currently pursuing a highly interdisiciplinary M.Sc. at University of Heidelberg, where I focus on applications of AIML in social and cognitive sciences. At the same time, I am pursuing the Certificate of Advanced Studies in Cognitive Science, a complementary interdisciplinary program to my degree, where I focus on aspects of Cognitive Science such as Computational Cognitive Science, fMRI Analysis, Linguistics, Philosophy and Behavioral Modeling.

For the past year I have been working as an intern in the Machine Learning Human Behavior research group, lead by Prof. Koppe, where I deal (as the name suggests) with Human Behavioral Modeling and Behavior Forecasting in the context of Repetead Trust Games. To this end I use both traditional statistical methods (such as Bayesian hierarchical modeling, Hidden Markov Chains etc.) and advanced Dynamical System Reconstruction models (such as shPLRNNs or AL-RNNs), with the goal of developing control mechanisms that can drive the agent behavior into desired equilibrium trust states (which I hope can contribute to individualized therapeutical applications or psychopathology). On the second half of my internship I have worked on adding Linear Control to the AL-RNN and evaluating it on classical control benchmarks from OpenAI Gym.

After a 6 months research internship at Roche Diagnostics, where I worked on Multivariate Time Series forecasting and modeling, this time with the aim of improving patient care through personalized predictions from wearables data, I am finally returning to academia with more pragmatism and more confidence to follow that which is close to me: understanding mental ilness through a computational framework. My own failed past experiences with psychiatry (both in Germany and abroad) showed me how big the methodological gap is, between the recent advances in AI and the understanding of cognitive dynamics, and somehow it dawned on me just now that I can pursue a career in this area. It may be a selfish pursuit, but at least I could shorten the gap through this selfishnes.

In my free time I enjoy philosophy, psychology, literature and art.

Scan me to go directly to the CV