Principal Research Scientist

I received my B.S. and Ph.D. degrees from Mississippi State University, both with distinctions, in 2007 and 2012, respectively. I conducted my dissertation work under the direction of Prof. James E. Fowler on the application of compressed sensing techniques to high-dimensional media, e.g. natural video, multiview image and video, light-field images, and hyperspectral imagery for geosciences. During my Ph.D. studies, I also worked as a research intern for Canon USA, Inc.'s research division in San Jose, CA, where my work lead to a patent for a novel lightfield acquisition system, for which I am listed as a co-inventor. After finishing my graduate studies, in 2013 I started a postdoctoral research position at Ecole Normale Supérieure (ENS) in the SPHINX group of Prof. Florent Krzakala. During this fruitful post-doc, I have focused my work mainly on the intersection of statistical physics and information problems, specifically, signal processing and machine learning. During my postdoc, I have been an author to a number of works on this subject, including works presented at both ICML and NIPS, and edited a notable book on the same topic. I have also been an invited speaker and lecturer at a number of venues across Europe to speak on the topic of statistical signal processing from the lens of statistical physics. In the second half of this postdoc, and during my subsequent work at Inria, I primarily focused on the topic of unsupervised machine learning; specifically, on how the study of statistical interpretations of neural models can lead to both practical and fundamental theoretical advances in machine learning. Since 2017, I have served as a Senior Machine Learning Scientist at Owkin, with a primary focus on privacy-aware machine learning and federated & distributed learning. From 2018, I have lead the Federated Learning Research Group at Owkin, where I manage a team of scientists and engineers conducting algorithmic research, performing federated data science, and developing software for federated learning on medical data. For more details, please visit my personal site: http://eric-tramel.github.io.