Brought to you by NumFOCUS Foundation and O’Reilly Media
The official Jupyter Conference
Aug 21-22, 2018: Training
Aug 22-24, 2018: Tutorials & Conference
New York, NY

Deep learning beyond the learning

Who is this presentation for?

Data Scientists, Developer, and Architects interested in building deep learning pipelines.

Prerequisite knowledge

Basic understanding of Data Science and Deep Learning is recommended.

What you'll learn

A deep learning pipeline is much more than just training. This talk presents the different challenges involved, and presents one potential end-to-end implementation.


Open Source frameworks such as Spark, TensorFlow, MXNet, and PyTorch enable anyone to model and train deep learning models. While there are many great tutorials and talks showing us the best ways for training models, there is little information on what happens before and after we have trained our model. How can we develop, store, utilize, test, and refine it?

In this talk we look at the complete deep learning pipeline and answer questions such as:

  • How can we enable data scientists to exploratively develop models without having to worry about the underlying infrastructure?
  • How can we easily deploy these distributed deep learning frameworks on any public or private infrastructure?
  • How can we manage multiple different deep learning frameworks on a single cluster, especially considering heterogeneous resources such as GPU?
  • What is the best interface for data scientist to use when working with the cluster?
  • How can we store and serve models at scale?
  • How can we update models which are currently in use without causing downtime for the service(s) using them?
  • How can we monitor the entire pipeline and track performance of the deployed models?