I am a Distinguished Research Scientist at Meta AI, working on the next generation of foundational generative AI models (Llama). Prior, I led a team of 130+ Meta AI researchers who developed technology to solve grand challenges in the natural sciences. My team delivered research breakthroughs including Cicero, ESM-2 and ESMFold, Llama, and OC22. I am currently building the next generation of generative AI models with the Llama team.
Before joining Meta, I worked as an Assistant Professor (with tenure) at Delft University of Technology, and as a post-doctoral researcher under supervision of Lawrence Saul at University of California, San Diego. I received my Ph.D. from Tilburg University in 2009 under the supervision of Eric Postma and Jaap van den Herik. I have visited the labs of Geoffrey Hinton at University of Toronto, and of Maja Pantic at Imperial College London. I obtained an M.Sc. degree in computer science from Maastricht University in 2005.
My main research interests are in artificial intelligence. In particular, I am interested in embedding models, large-scale weakly supervised learning, visual reasoning, cost-sensitive learning, language modeling, and applications in the arts and sciences.
Press and Awards
My work on web-scale weakly supervised learning was featured in Meta CTO Mike Schroepfer’s keynote at F8 2018 (at 14:27). The work also received extensive press coverage in, among others, Wired, VentureBeat, The Verge, and ZDNet.
Our DenseNet paper received the CVPR 2017 Best Paper Award. The Fisher information loss paper received the UAI 2021 Best Paper Award. Prior, I received the SNN Machine Learning Award 2012 and won the Merck Visualization Challenge on Kaggle.