The Fabula team at Twitter UK office. Left to right: Michael Bronstein, Fabrizio Frasca, Emanuele Rossi, Davide Eynard, Federico Monti, Damon Mannion (image: Stacey Conti).
On June 3, 2019, Twitter announced the acquisition of Fabula AI, a company founded by a team of USI researchers headed by Prof. Michael Bronstein. As a result of the acquisition, Bronstein joins Twitter as its new Head of Graph Learning Research. Also joining Bronstein are USI researchers Federico Monti, Davide Eynard and Fabrizio Frasca. Monti and Frasca were PhD students in Bronstein’s group, and Eynard was his postdoc. Parag Agrawal, CTO at Twitter, announced the details of the acquisition in a Twitter blog post.
"I am truly privileged to have worked with such a stellar team at USI. This acquisition is the culmination of the hard work of all my current and former students," said Bronstein, who was the founder and Chief Scientist of Fabula.
According to Federico Monti, founder and former CTO of Fabula, "There is no better satisfaction than realizing that the methods we have been developing for several years will be used on a service like Twitter to ensure people feel safe and see relevant information."
Monti did his undergraduate studies at Politecnico di Milano before joining Bronstein’s group in 2016. He pioneered some of the first methods for deep learning on graphs, leveraged by Fabula.
"Federico is one of the best students I have worked with in my career, a true superstar," says Bronstein. "He has published papers already regarded as seminal in top conferences and journals, and the acquisition of Fabula now crowns his work. It is hard to think of anything more rewarding for a professor than the success of his pupils."
Fabula AI was founded by Michael Bronstein (Chief Scientist), Federico Monti (CTO), Damon Mannion (COO) and Ernesto Schmitt (President). The company was built upon the pioneering research that Prof. Bronstein and his team conducted at USI in the domain of geometric deep learning, a novel class of powerful ML techniques applicable to network-structured data. The result is the ability to analyze very large and complex datasets describing relations and interactions, and to extract signals in ways that traditional ML techniques are not capable of doing. Bronstein and his collaborators have previously successfully applied geometric deep learning to a broad range of problems ranging from protein design [ www.biorxiv.org/content/10.1101/606202v1 ] to neutrino detection [ www.usi.ch/en/feeds/9771 ].
Bronstein joined the Institute of Computational Science ( ICS ) at the USI Faculty of Informatics in 2010. During his tenure at USI, he was awarded four ERC grants, considered the most prestigious and highly competitive research funding in Europe - a Starting grant in 2012, a Consolidator grant in 2016, and two Proof of Concept grants in 2016 and 2018. Besides his successful academic career, which has been recognized by numerous accolades (including election to Fellow of IEEE [ www.inf.usi.ch/en/feeds/9553 ], one of the highest recognitions for an engineer, in 2018), Bronstein is an accomplished serial entrepreneur. Fabula is already his fourth startup and second successful exit; in fact, he has previously worked with his fellow co-founder Ernesto Schmitt at Invision that was acquired by Intel in 2012 and led to the successful RealSense 3D sensing technology.
"I think there is an enormous pool of talent and potential at the University. However, there is a big gap between publishing a good paper, developing good technology, and buildingmaking a successful company. Each of these steps is at least an order of magnitude harder, and requires a different mindset. One of the challenges of commercializing academic research is that as academics, we are centered on our research and then look where to apply it - it’s a solution looking for a problem. To make a successful company, you need to think the other way around," says Bronstein.
Second, Bronstein notes that, "many academics are not aware of the commercial implications of publishing their results or code, such as eligibility for patent protection."
Fortunately, USI has an excellent Research Support unit with professional and knowledgeable staff that can advise on the right steps to protect inventions. Being a small university with lean administration is also instrumental in obtaining technology licensing agreements without having to deal with too much red tape. "I have only superlatives to say about Andrea Foglia and all the technology transfer office staff," says Bronstein.
Finally, there is also an element of luck and being in the right place at the right time.
"I have seen many great teams doing fine projects and yet struggling to become commercially successful," says Bronstein. "So serendipity is definitely an important ingredient of success. However, there is also a saying that, ’luck comes to those with brains.’"