University of Cambridge / Adap GmbH
Cambridge, UK / Hamburg, DE
Daniel J. Beutel developed a novel Federated Learning framework named "Flower" (https://flower.dev) based on his prior work at the University of Oxford. Flower enables everyone to build federated machine learning systems that move AI training to edge devices (instead of moving the data to the cloud). It's already being used by some of the best researchers in the world, for example, for quantifying the carbon impact of Federated Learning (NeurIPS 2020 Workshop "Tackling Climate Change with Machine Learning"). In addition, there are the first industrial users who are evaluating Flower in privacy-relevant areas such as healthcare or video analysis on edge devices. At the beginning of 2020 Daniel (together with Taner Topal) founded the startup Adap based in Hamburg. Adap accelerates the transfer of novel AI approaches from research to production usage and is the leading developer behind Flower. One of its first customers is the University of Cambridge, which adopted Flower in teaching.
How can we improve the carbon impact of our machine learning systems? How can we scale up our federated learning research systems to make the results more applicable to real-world scenarios? What's the impact of systems heterogeneity on federated learning?