Otniel-Bogdan Mercea

I am on the job market looking for research scientist roles. If you have a project that fits with my research skills, do not hesitate to contact me.

I am a final-year PhD candidate jointly affiliated with the Max Planck Institute for Intelligent Systems and the University of Tubingen, part of the IMPRS-IS doctoral program. My research is supervised by Prof. Zeynep Akata and Prof. Andreas Geiger. During my PhD, I was also a guest PhD student at Helmholtz Munich and Technical University of Munich, supervised by Prof. Zeynep Akata.

I earned my MSc in Artificial Intelligence from the University of Edinburgh in 2020, completing my thesis under the supervision of Prof. Amos Storkey. Prior to that, I obtained my BEng in Computers and Information technology from Politehnica University of Timisoara in 2019, with my undergraduate thesis supervised by Prof. Calin-Adrian Popa.

I have had the privilege of applying my research expertise in industry through several impactful opportunities. In the autumn of 2024, I interned at Google DeepMind, where I worked under the supervision of Stefano Pellegrini, Jasper Uijlings, and Cordelia Schmid on enhancing SAM 2 for scenarios involving significant occlusion or movement. Previously, from 2023 to 2024, I spent eight months at Google Research, collaborating with Anurag Arnab, Alexey Gritsenko, and Cordelia Schmid on the efficient adaptation of large-scale models. During this time, I also worked with Aleksandra Nowak, Utku Evci, and Yann Dauphin from Google DeepMind on a related project in efficient adaptation strategies. Prior to these roles, I was a Machine Learning Researcher at Everseen, where I focused on real-time multi-camera tracking systems.

Most of my research focuses on improving deep learning efficiency through data-efficient (low-shot) and model-efficient (adaptation) methods for large-scale models. Furthermore, I have extensive experience in multimodal learning, including audio-visual learning and multi-modal large language models, as well as in video semantic segmentation and tracking.