Presented By O’Reilly and Intel AI
Put AI to Work
April 29-30, 2018: Training
April 30-May 2, 2018: Tutorials & Conference
New York, NY

Deploy MXNet and TensorFlow deep learning models with AWS Lambda, Google Cloud Functions, and Azure Functions

Greg Werner (3Blades), N C
1:40pm–5:10pm Monday, April 30, 2018
Average rating: ****.
(4.00, 2 ratings)

Who is this presentation for?

  • Anyone looking for increased efficiencies for their data science teams efforts

Prerequisite knowledge

  • A working knowledge of the TensorFlow and MXNet deep learning frameworks and Python
  • A basic understanding of development and operations concepts (i.e., DevOps), such as continuous integration and continuous deployment

Materials or downloads needed in advance

  • If you'd like to follow along, you'll need a WiFi-enabled laptop with the latest version of Docker installed, along with AWS, Google Cloud, and Azure accounts.

What you'll learn

  • Learn how to deploy trained models using TensorFlow and MXNet along with the leading serverless solutions in the market, how to surface TensorFlow and MXNet as RESTful endpoints using performant and cost-effective solutions, and how to collect data from deployed models for monitoring and data collection


There’s an extensive amount of literature about how to use the various available machine and deep learning frameworks to train, save, and optimize models, most of which is extremely detailed and very focused on the programming and math aspects of the process—making it relatively easy to get up and running. However, it’s also important to consider deployment alternatives so models may be used in staging and production environments using performant, secure solutions.

Greg Werner walks you through using MXNet and TensorFlow to train deep learning models and deploy them using the leading serverless compute services in the market: AWS Lambda, Google Cloud Functions, and Azure Functions. You’ll also learn how to monitor and iterate upon trained models for continued success using standard development and operations tools. Examples will demonstrate alternatives for automating the data science pipeline using continuous integration and continuous deployment (CI/CD) tools to reduce human error, leverage battle-tested cloud infrastructure solutions, and reduce opportunity costs by quickly pushing the most effective models to production quickly and securely.

Photo of Greg Werner

Greg Werner


Greg Werner is the founder and CEO of 3Blades. 3Blades develops and maintains IllumiDesk ( and Cup of Data ( Greg has built information technology businesses his entire career. Previously, he cofounded Certsuperior, currently one of the largest web security companies by sales in Latin America, and Reachcore, a leading business-to-business supplier of document exchange services for the oil and gas, insurance, telco, and financial verticals. Greg is a co-organizer of the PyData Meetup group in Atlanta. He frequently contributes to open source projects that help the scientific community, particularly those within the Python ecosystem. Greg holds a BA in economics from Emory University, an MBA in international management from Thunderbird, and a master’s degree in computer science from the University of Illinois.


Comments on this page are now closed.


Picture of Greg Werner
Greg Werner | CEO
04/30/2018 9:41am EDT

Picture of Greg Werner
Greg Werner | CEO
04/30/2018 3:49am EDT

Hi Thomas, we provided some instructions in this repo: We are still merging some of the labs but the Readme has the setup instructions. We will do some research this morning to see who we could help you test with these vendors without having to pay for it.

thomas gebregergis | DATA SCIENTIST
04/30/2018 3:03am EDT

Hello Greg, any chance you can share the download links for AWS, Google Cloud and Azure if they have free versions for the sake of this tutorial? thanks