In the Research Spotlight: Mu Li

As AWS continues to support the Artificial Intelligence (AI) community with contributions to Apache MXNet and the release of Amazon Lex, Amazon Polly, and Amazon Rekognition managed services, we are also expanding our team of AI experts, who have one primary mission: To lower the barrier to AI for all AWS developers, making AI more accessible and easy to use. As Swami Sivasubramanian, VP of Machine Learning at AWS, succinctly stated, “We want to democratize AI.”

In our Research Spotlight series, I spend some time with these AI team members for in-depth conversations about their experiences and get a peek into what they’re working on at AWS.

Mu Li is a principal scientist for machine learning at AWS. Before joining AWS, he was the CTO of Marianas Labs, an AI start-up. He also served as a principal research architect at the Institute of Deep Learning at Baidu. He obtained his PhD in computer science from Carnegie Mellon University, where one of his advisors was Alex Smola, now Director of Machine Learning at AWS. Mu’s research has focused on large-scale machine learning. In particular, he is interested in the co-design of distributed systems and machine learning algorithms. He has been the first-author for computer science conference and journal papers on subjects that span theory (FOCS), machine learning (NIPS, ICML), applications (CVPR, KDD), and operating systems (OSDI).

At AWS, Mu leads a team that works primarily on the Apache MXNet framework. Their focus is making it easier to use deep learning and to run deep learning applications on AWS. To accomplish this, Mu and his team are charting new territory in deep learning research, investigating and simplifying new algorithms that can run on large-scale datasets in distributed systems. “The speed of machine learning training depends on two things: how fast you can process images and how fast you can process the final model,” says Mu. The framework should support using multiple GPUs and multiple machines. “The latter is related to optimization–what we call the convergence rate. When we move from a single machine to multiple machines, we need to develop new distributed training algorithms. We need to change the algorithm itself–to change the neural network structure–so that it can be easily used to train very large datasets on a large number of machines.”

In essence, Mu is ensuring that the system fits the algorithms and is changing the algorithms to fit the systems. “My vision is to train a model with one click and deploy a model with one click.” That is, to accelerate the use of deep learning on the AWS platform by removing all of the complexity for users and reducing training time. As Mu says, “Pushing state-of-the-art deep learning to a large scale and making the technology available to every AWS user–this is the goal.”

When not solving deep learning problems at AWS, Mu enjoys spending time with his family.  And as a new father of a 4-month old, Mu is very busy when outside of AWS!

Victoria Kouyoumjian is a Sr. Product Marketing Manager for the AWS AI portfolio of services which includes Amazon Lex, Amazon Polly, and Amazon Rekognition, as well the AWS marketing initiatives with Apache MXNet. She lives in Southern California on an avocado farm and can’t wait until AI can clone her.