Scalable training and serving of personalized models

The past decade has seen substantial growth in Learning Systems research [1, 9, 12, 14, 22] combining advances in system design with new efficient algorithms to enable the training of complex models on vast amounts of data. As a consequence we have seen widespread adoption of machine learning techniques to address important real-world problems [5, 17].While this work has been wildly successful in shaping both the machine learning [19, 16, 11] and systems fields [14, 15, 1], it also ignores a big part of real-world machine learning. In particular, much of the work in Learning Systems has operated under the fiction: the world hands me a static, potentially very large, dataset and I train an accurate, potentially complex, model. This fiction departs from reality in two key regards that we begin to address in this work.