Marin Vlastelica
- PostDoc
- marin.vlastelica@ai.ethz.ch
- Personal Website
- I am a firm believer that we can utilize large models in order to discover new knowledge, with great applications in science and the creative arts. Therefore, I am interested in principled discovery and (constrained) optimization methods via generative foundation models in different domains (such as image, language, robots and further). This involves topics such as efficient finetuning methods, uncertainty quantification, reinforcement learning, sampling procedures and further.