Put AI to work
June 26-27, 2017: Training
June 27-29, 2017: Tutorials & Conference
New York, NY

Distributed deep learning on AWS using Apache MXNet

Joseph Spisak (Amazon), Sunil Mallya (Amazon Web Services)
9:00am12:30pm Tuesday, June 27, 2017
Implementing AI
Location: Sutton Center Level: Intermediate
Secondary topics:  Cloud, Deep Learning

Prerequisite Knowledge

  • Familiarity with Python and the Jupyter Notebook
  • Experience with machine learning and data science

Materials or downloads needed in advance

What you'll learn

  • Gain a foundational understanding of deep learning, Amazon AI services, and Amazon’s strategy for enabling developers
  • Learn how to set up a GPU instance or a secure cluster of instances on AWS for distributed model training using AWS’s free deep learning tools
  • Learn how to use Apache MXNet to ingest data, train models to use multiple GPUs, fine-tune pretrained models (transfer learning), and deploy on services such as AWS Lambda to serve in a production environment


Deep learning continues to push the state of the art in domains such as computer vision, natural language understanding, and recommendation engines. One of the key reasons for this progress is the availability of highly flexible and developer-friendly deep learning frameworks. Joseph Spisak and Sunil Mallya offer an introduction to the powerful and scalable deep learning framework Apache MXNet. You’ll gain hands-on experience using MXNet with preconfigured Deep Learning AMIs and CloudFormation Templates to help speed your development and quickly spin up AWS GPU clusters to train at record speeds. You’ll also learn how to leverage the NVIDIA DLI as you walk through the mechanics of using MXNet and explore several end-to-end application examples that apply deep learning to event prediction, language modeling, and computer vision.

Topics include:

  • An overview of Amazon AI and a brief background on deep learning
  • Getting started quickly on AWS using Deep Learning AMIs with one to hundreds of GPUs
  • Using MXNet’s imperative interface (NDArray), symbolic interface (Symbols) and module API to train and predict
  • Hands-on with Apache MXNet: Application examples targeting time series predictions using LSTMs, object detection using faster RCNN, and fine-tuning a state-of-the-art CNN
Photo of Joseph Spisak

Joseph Spisak


Joseph Spisak manages deep learning product management at AWS. Joseph has experience driving strategies and technical and business engagements around machine learning-based cloud workloads, such as computer vision, natural language processing, video summarization and analysis, and speech recognition. Joseph has more than 15 years’ experience delivering products and services in digital video, cloud-based media transcoding, image processing, and machine and deep learning in the consumer mobile, broadcast, and cloud segments. Joseph holds a bachelor’s degree in electrical engineering from Michigan State University and an MBA and MS in finance from the University of Denver. He is a proud graduate of the Entrepreneurial and Innovation certificate program at Stanford University’s Graduate School of Business.

Photo of Sunil Mallya

Sunil Mallya

Amazon Web Services

Sunil Mallya is a solutions architect focused on deep learning at AWS, where he works with customers in various industry verticals. Sunil has an acute passion for serverless computing. Previously, he cofounded the neuroscience- and machine learning-based image analysis and video thumbnail recommendation company Neon labs and worked on building large-scale low-latency systems at Zynga. He hold a master’s degree in computer science from Brown University.