Embedded Files
Bio
Adriana Romero-Soriano is a Research Scientist at Meta FAIR (Fundamental AI Research), an Adjunct Professor at McGill University's School of Computer Science, a core academic member of Mila (Quebec Artificial Intelligence Institute), and the recipient of a prestigious Canada CIFAR AI Chair. Adriana’s research is situated at the intersection of computer vision, generative modeling, and representation learning. She is recognized for her contributions to image and video generation, representation learning and active sensing. Her most recent work pushes the frontiers of vision and multi-modal generative models by devising novel model sampling, search, and post-training techniques to unlock high-utility and physically plausible synthetic data from generative models.
Adriana is an active member of the AI community: she has co-organized numerous workshops at top-tier conferences, including the ReGenAI workshop at CVPR (2024-2025) and sessions on the Science and Engineering of Deep Learning and Graph Representation Learning at NeurIPS and ICLR. Adriana earned her Ph.D. from the University of Barcelona under the supervision of Dr. Carlo Gatta, followed by a postdoctoral fellowship at Mila (Québec AI Institute) with Prof. Yoshua Bengio, where she advanced deep learning for complex biomedical and graph-structured data.
Pro bono work
Beyond her scientific contributions, Adriana has served as the president of the management board of a small non-profit (25-30 employees, and 100 members) dedicated to early childhood care and development for a two-year period. In this role, she managed the organization and successfully led its organizational restructuring.
Research highlights
Vision synthetic data: Co-engineered a synthetic training data research program that substantially improved the utility of synthetic training images, enabling more effective model training through tailored synthetic data curriculums.
Vision generative models: Developed a state-of-the-art image generative model to facilitate critical research on sample representation diversity, co-leading a team of 10+ researchers.
Physics in video generation: Pioneered inference-time physics alignment for video generative models, achieving over 6% improvement over prior state-of-the-art. This contribution resulted in the winning entry of the Physics IQ challenge hosted during the third perception test challenge at ICCV'25.
AI & healthcare: Contributed to the high-profile fastMRI project, which enabled the acceleration of MRI acquisition by 2x and transitioned from research to clinical practice. Co-led seminal work on active acquisition methods to improve MRI reconstruction efficiency.
Recipe generation: Co-led the development of the first image-to-recipe system, which was featured in Medium and Neurohive, and which resulted in an research-to-demo showcased at CVPR'19.
Graph representations: Co-authored seminal work on graph attention networks, presented at ICLR'18.
Google Sites
Report abuse