Petar Veličković

Petar Veličković is a Senior Staff Research Scientist at Google DeepMind and an Affiliated Lecturer at the University of Cambridge. He is also an Associate of Clare Hall, a graduate college within the University of Cambridge.

• PhD in Computer Science from the University of Cambridge (Trinity College), supervised by Pietro Liò.

His research focuses on aligning neural networks with classical computation to enhance out-of-distribution generalization. Key areas include:

• Neural Algorithmic Reasoning (assessing algorithm execution in neural networks)
• Graph Representation Learning (enhancing graph-based models)
• Geometric Deep Learning (co-authored a proto-book)

His recent publications span conferences like LoG and NeurIPS, emphasizing algorithmic robustness and graph neural networks. Notable work includes Graph Attention Networks and Deep Graph Infomax, with applications in Google Maps and AI-assisted mathematical discovery.

Scientific accolades include:

• Best Paper Award for 'Expander Graph Propagation' at NeurIPS'22 GLFrontiers
• ELLIS Scholar in Geometric Deep Learning Program

He has advised competitive programming teams at Cambridge and contributed to integrating AI with classical computational frameworks. His work has been featured in major outlets like Nature, Science, and The Verge.