Put open source to work
July 16–17, 2018: Training & Tutorials
July 18–19, 2018: Conference
Portland, OR

Edge ML: Deep learning on IoT devices

Matt Ellis (TIBCO Software), Rei Kurokawa (Hitachi High-Tech Solutions)
1:45pm2:25pm Thursday, July 19, 2018
Edge computing, TensorFlow
Location: E146
Tags: tensorflow
Level: Intermediate
Average rating: *****
(5.00, 1 rating)

Who is this presentation for?

  • Enterprise developers and managers

Prerequisite knowledge

  • Familiarity with machine learning concepts and edge computing

What you'll learn

  • Learn how edge ML can lead to significant savings in operational costs (up to 85% in some cases)
  • Explore data science techniques for time-based edge data
  • Discover an approach using TensorFlow and Project Flogo to make predictions directly on edge devices without depending on cloud computing


A single device can generate thousands of records of data every second. In traditional implementations, all data is transmitted back to a server or gateway for scoring by a machine learning model. This data is also stored in a data repository for later use by data analysts and scientists. By the year 2020, the world will have an estimated 20 billion IoT devices. Storing, processing, reasoning with, and extracting business value out of this data will require huge computational and financial resources.

Matt Ellis and Rei Kurokawa share an approach that draws on big data and deep learning to make predictions directly on edge devices without depending on cloud computing. Matt and Rei demonstrate deployments of Google TensorFlow models directly onto edge devices or into serverless architectures using Project Flogo, a lightweight open source edge microservices framework that enables you to build applications that run on edge devices and integrate them with IoT gateways as well as extend the reach of core applications and infrastructure to edge devices to interconnect everything anywhere. They conclude by discussing industry-specific use cases and scenarios and explaining how reducing ingress traffic to cloud services can lead to an 80% cost savings.

Photo of Matt Ellis

Matt Ellis

TIBCO Software

Matt Ellis is a senior product manager and head of open source software at TIBCO Software, where he focuses on product and strategy around open source and Project Flogo, an open source edge microservices framework built entirely in Go. As of late, Matt has spent his time focusing on two key technical shifts: machine learning and serverless compute. Matt has brought many technical achievements to the Golang open source community in areas related to edge machine learning with TensorFlow and serverless compute constructs in Golang with AWS Lambda. Matt has been involved in the tech industry for over two decades and has held a number of roles. Early in his career, he developed 3D rendering engines in both OpenGL and Direct3D and authored a number of technical manuals and books on the topic. During an independent consulting opportunity in Brazil, Matt authored two additional books focusing on RESTful services and APIs.

Photo of Rei Kurokawa

Rei Kurokawa

Hitachi High-Tech Solutions

Rei Kurokawa is an IoT and cloud solutions manager at the Tokyo office of Hitachi High-Tech Solutions. Rei’s expertise includes ERP systems, cybersecurity systems, cloud-based solutions, data center operation management, and industrial IoT solutions. She has worked in a variety of industries in both the private and public sectors, including infrastructure, consumer services, finance, and the chemical, pharmaceutical, and manufacturing industries. She frequently advises company executives and their counsel on matters related to cybersecurity and BCP management as well as generating benefits by implementing IoT solutions such as edge computing.