Distributed Deep Learning on Azure Databricks

Deep learning algorithms have helped accelerate AI achievements within the recent years, particularly in the fields of computer vision, speech recognition, and natural language processing. Democratization of deep learning libraries and frameworks enable data scientists and machine learning engineers to apply state-of-the-art algorithms to solve high-valued business challenges. However, it is not trivial to stand up such a development environment with GPU and necessary libraries installed.

In this presentation, we will discuss how to quickly stand up a deep learning experiment environment on Azure Databricks using the Databricks Runtime for ML. It includes native support for popular Python deep learning frameworks such as TensorFlow and Keras, as well as OpenCV and PIL for computer vision and image processing. This runtime enables an immediate environment to overcome some common challenges practitioners face in getting started with deep learning. In addition, we will show how to easily set up distributed image processing and model training with Azure Databricks.

Agenda:
6:00-6:30 pm: Networking & Pizza
6:30-7:00 pm: Kickoff and Overview of Databricks by Samir Gupta, Solution Architect, Databricks
7:00-8:00 pm: Distributed Deep Learning Talk & Demo by Hubert Duan, Data and AI Specialist, Microsoft