Professor Ann Copestake, Head of the Department of Computer Science and Technology says “Machine learning is a vital area of research, where connections with industry partners can drive advances in understanding. We are incredibly grateful for Twitter’s support of these new postgraduate scholarships. The recipients this year will be working on deep learning and AI safety, and on machine learning over graph-structured data.”

Dr Ferenc Huszár, a Senior Lecturer in Machine Learning, is also a staff research scientist at Twitter, advising Twitter's ML Ethics, Transparency and Accountability (META) team. He comments: "Cambridge has a strong tradition in fundamental ML research and our Department has particular strengths in NLP, graph machine learning, responsible ML and AI policy. These areas are aligned with the focus of Twitter's applied research programmes. This partnership is bound to create amazing pathways for real-world impact for our basic research."

He adds: "These scholarships enable us to fund the most promising future researchers and they contribute to our goals of widening participation in machine learning research. This is particularly important in responsible AI where we hope the scholarships will attract applicants who may not have previously considered a degree in Advanced Computer Science."

Pietro Liò, Professor of Computational Biology at the University of Cambridge, says: "There are already strong existing links between Twitter and the University of Cambridge including through research and student internships. The connection with Twitter also benefits us in understanding better how to build AI for social good and combat fake news that can have a damaging impact on society. The new Twitter Scholarships will reinforce these links and enable the Twitter PhD students to investigate cool areas of graph representation, topological and geometrical properties of information processing.” 

“It’s a great privilege for me to be one of the first students to receive a Twitter Scholarship,” says Iulia Duta, one of the first two Twitter PhD Students, who graduated from the University of Bucharest as the top computer science student of her cohort in 2018. “The support this studentship offers is making it possible for me to start my doctoral studies this year, and I couldn't be more grateful for this opportunity.” 

Since graduating, Iulia has been working as a machine learning researcher at Bitdefender. She has a keen interest in relational machine learning and graph neural networks and has co-authored multiple papers on the topic, including one presented at NeurIPS 2019. In her PhD she will focus on graph neural networks, supervised by Professor Pietro Liò with Dr Ferenc Huszár as a secondary supervisor. 

She adds: “After a few years working as a machine learning researcher, a PhD seemed the perfect next step for me but I have struggled with a lack of funding. I am excited to have this opportunity now to grow professionally in a group with great knowledge and experience in the field. I hope my PhD will allow me to express my curiosity about relational models and beyond, gain in-depth research knowledge and skills and contribute further to this area of research.” 

“I am very grateful to have been provided this scholarship from Twitter to pursue a PhD in machine learning at Cambridge,” says Nitarshan Rajkumar, currently an MSc student at the Mila lab in Montreal. “As an international student, the costs of pursuing postgraduate study would otherwise be prohibitive. It is also exciting to see increased investment being placed into research of this transformative technology at the department.” 

Nitarshan has spent several years in industry as a software engineer at Airbnb. His interests include understanding how deep learning works and the responsible development of AI. He has co-authored multiple papers, including one presented at NeurIPS. Nitarshan will work under the supervision of Dr Ferenc Huszár with Dr David Krueger as a secondary advisor on AI safety. 

He adds: “Deep learning has seen enormous success in recent years, yet progress has largely been driven by intuition rather than a principled understanding of how neural networks work, making it difficult to reason about increasingly complex models in a systematic manner. Accordingly, my research focus will be on empirical approaches to identifying and characterizing fundamental properties of deep learning such as generalisation, scalability, and robustness. I also hope to contribute to societal aspects of AI safety such as effective governance and policy mechanisms."