Pedro Bizarro explains the origins, strengths and main results of the CAMELOT project
As part of the CAMELOT project, promoted by Feedzai and co-financed by COMPETE 2020, we interviewed Pedro Bizarro, the company’s Chief Science Officer, who told us about the strategic vision, the main milestones and the legacy of this initiative. CAMELOT stood out for its ambition to develop a new generation of machine learning solutions, tackling some of the most demanding challenges of modern artificial intelligence in an innovative way. In this conversation, Pedro Bizarro shares with us how the project came about, what makes it truly distinctive, the results achieved along the way and the lasting impact of close collaboration between academia and industry.
How did the CAMELOT project come about? What were the main motivations?
CAMELOT came about by harnessing the complementary skills of teams from Feedzai, the University of Coimbra, the Faculty of Sciences of the University of Lisbon, the Instituto Superior Técnico and Carnegie Mellon University: the common goal was to create a truly innovative machine learning platform capable of responding to three major challenges facing modern artificial intelligence. Firstly, it was essential to guarantee real-time response times during both the training and inference phases of the models, while keeping cloud operating costs to a minimum. Secondly, the aim was to enable training on previously anonymised data, overcoming the privacy restrictions associated with sharing information between different entities. Finally, it was crucial to simplify the integration and synchronisation of multiple data platforms – from relational databases to graphs or key-value stores – using a Domain Specific Language (DSL) that would abstract away all this complexity.
What do you consider to be the differentiating element of the project?
The differentiating element of the project was that it addressed real problems with great impact, which are often ignored by the state of the art because they are so difficult to solve.
What are the most important results to highlight?
So far, the project has generated 35 scientific publications accepted at conferences or workshops and has been invited to present five tutorials or keynotes at international events. It has also produced a book chapter and supported nine completed master’s dissertations, as well as having had the additional collaboration of four other master’s students and five doctoral students. In the field of intellectual property, it resulted in three provisional patents and one patent already granted. Finally, the work was honoured with three awards.
If you had to summarise the legacy of the CAMELOT project in a few words, which would you choose?
The legacy of the project has been to create a collaborative scientific climate focussed on solving very difficult problems with high quality by taking advantage of young talent.
Through the CAMELOT project, Feedzai and its partners have made a solid contribution to the advancement of artificial intelligence, demonstrating how collaboration between science and industry can generate innovation with real and lasting impact.