Due to increased model complexity and training data set size, the computational cost of training and deploying today’s state-of -the-art ML systems has grown rapidly. The MADLab team is developing new hardware and software solutions to make ML more scalable and computationally efficient.