Keynote Speakers

Pauline Bolignano

Pauline Bolignano is the manager of the Automated Reasoning team at Prime Video, whose main focus is to provide Prime Video developers with automated reasoning-based tools to assist them in confidently evolving the App. Prior to that, she worked as an Applied Scientist in the Prime Video Automated Reasoning team and in the AWS Automated Reasoning team. She received her PhD from the University of Rennes, France, in 2017, where she worked on verification of security properties of operating systems.

Title of the Talk:

How do we do Science at Prime Video?


Why do we do science and how do we do it? By taking some concrete examples of problems we are solving at Prime Video, I’ll give you some insights about how we think about a science problem and how we tackle it. I’ll take my examples from various scientific domains: Automated Reasoning, Anomaly Detection and Computer Vision. This talk will also show what a PhD internship with us may look like.

David Pfau

David Pfau is a staff research scientist at DeepMind. He's also a visiting professor at Imperial College London in the Department of Physics, where he supervises work on applications of deep learning to computational quantum mechanics. His research interests span artificial intelligence, machine learning and scientific computing. Prior to joining DeepMind, he was a PhD student at the Center for Neuroscience at Columbia. His current research interests include applications of machine learning to computational physics and connections between differential geometry and unsupervised learning.

Title of the Talk: Deep Learning and Ab-initio Quantum Chemistry and Materials


The deep learning revolution has led to enormous advances in artificial intelligence, and significant interest in bringing these advances to the physical science as well. Much of the effort in deep learning for chemistry has focused on data-driven problems, such as generative models of molecules, or supervised learning of more accurate molecular force fields. Less attention has been given to ab-initio quantum methods, which seek to solve for the many-electron wavefunction directly from first principles. In this talk, I will show how deep learning can be used to bring unprecedented accuracy to a class of ab-initio methods called variational quantum Monte Carlo (VMC). Methods familiar from machine learning, such as natural gradient descent, MCMC and high dimensional function approximation, translate naturally to VMC. I will describe how we developed the Fermionic Neural Network (FermiNet) and used it to bring unprecedented accuracy to calculating the properties of molecules, as well as the uniform electron gas, a model system for materials. I will also introduce the Psiformer, a neural network architecture built around self-attention modules, similar to the celebrated Transformer, which reaches significantly higher accuracy on large molecular systems than the FermiNet. This is joint work between DeepMind and the group of Matthew Foulkes in the Department of Physics here at Imperial.

Thomas Barrett

Dr Thomas Barrett is a Research Scientist and Team Lead at InstaDeep working across fundamental machine learning research and applications in the natural sciences. A physicist by training, Tom spent several years (2014-2020) at the University of Oxford completing an (experimental) PhD and (theoretical) Post-Doc, before he joined InstaDeep in early 2021 to complete his escape from basement laser laboratories. His current research interests are the application of reinforcement learning to NP-hard combinatorial problems, developing and training the next generation of AlphaFold-inspired models for protein structure prediction, and setting up a new Quantum Machine Learning team.

Title of the Talk: On the role of Research in the industry