Presenters

Petar Veličković is a Staff Research Scientist at DeepMind, Affiliated Lecturer at the University of Cambridge, and an Associate of Clare Hall, Cambridge. Petar’s research concerns geometric deep learning—devising neural network architectures that respect the invariances and symmetries in data (a topic he’s co-written a proto-book about). Within this area, Petar focusses on graph representation learning and its applications in neural algorithmic reasoning. Petar’s research has been used in substantially improving the travel-time predictions in Google Maps, and guiding the intuition of mathematicians towards new top-tier theorems and conjectures. Petar has published several first-author papers in the area of neural algorithmic reasoning at top-tier venues (ICLR’20, NeurIPS’20 spotlight, ICML’22, NeurIPS’22), including the CLRS-30 benchmark, which is the first general-purpose unified algorithmic reasoning benchmark.

Andreea Deac is a PhD student at Université de Montréal and Mila, supervised by Jian Tang. Andreea is broadly interested in how learning can be improved through the use of graph representations, having worked on neural algorithmic reasoning, reinforcement learning and applications to biotechnology, in particular drug discovery. During her PhD, Andreea spent time at DeepMind, working with Doina Precup, Théophane Weber and George Papamakarios, and at MSR Cambridge, working with Marc Brockschmidt. Prior to that, Andreea completed her MEng with distinction under the supervision of Pietro Liò at University of Cambridge. Andreea has published several key contributions to neural algorithmic reasoning and its deployments in downstream tasks, including the eXecuted Latent Value Iteration Network (XLVIN, NeurIPS’21 spotlight), which is the first published evidence of downstream utility of neural algorithmic reasoning.

Andrew Dudzik is a Senior Research Engineer at DeepMind working on theoretical foundations of algorithmic alignment and GNNs, and making things run fast. Before that, he was fifth author on the AlphaStar paper, working in diverse areas like multi-agent and architecture innovation, as well as scaling and resource management. Before that, he got his PhD at U.C. Berkeley in hyperalgebra and algebraic geometry. And before that, he worked in finance, search, and even as an Agile software consultant. Andrew has published a joint-first-author paper on the categorical foundations of algorithmic reasoning (NeurIPS’22), which is the first paper attempting to thoroughly ground the principles of algorithmic alignment.