Presented By O'Reilly and Cloudera
Make Data Work
September 25–26, 2017: Training
September 26–28, 2017: Tutorials & Conference
New York, NY

GPU-accelerating a deep learning anomaly detection platform

Joshua Patterson (NVIDIA), Michael Balint (NVIDIA), Satish Varma Dandu (NVIDIA)
2:55pm3:35pm Wednesday, September 27, 2017
Data science & advanced analytics, Machine Learning & Data Science
Location: 1A 12/14 Level: Intermediate
Secondary topics:  Deep learning
Average rating: ****.
(4.00, 1 rating)

Who is this presentation for?

  • CTOs, CIOs, data engineers, and data scientists

Prerequisite knowledge

  • Basic knowledge of deep learning frameworks
  • Experience with massive-scale event logs and monitoring systems

What you'll learn

  • Explore how NVIDIA leverages its GPUs to accelerate anomaly detection
  • Learn about the challenges of training and deploying a deep learning model in a production environment and how to overcome them
  • Discover which GPU-accelerated technologies can be used to dramatically speed up a data pipeline


How can deep learning be employed to create a system that monitors network traffic, operations data, and system logs to reliably flag risk and unearth potential threats? There are many challenges with developing such a system. With new types of behavior constantly emerging, a robust and exhaustively tagged dataset is extremely difficult to obtain. Training a model on such a large amount of data can often take longer than practicality might dictate. Once trained, it is often difficult to employ the model in an expeditious way and get actionable results in a production environment.

Drawing on NVIDIA’s system for detecting anomalies on various NVIDIA platforms, Satish Dandu, Michael Balint, and Joshua Patterson explain how to bootstrap a deep learning framework to detect risk and threats in operational production systems, using best-of-breed GPU-accelerated open source tools. They then demonstrate how to speed up the training cycle by employing several algorithmic hacks and leveraging a cluster of GPUs, shortening the training time from days to hours, dramatically cutting the inferencing time, and generally making the entire system much more adaptive. Join Satish, Michael, and Josh as they walk you through how they built such a system, covering the architecture, the algorithms they implemented, how they sped up various parts of the data pipeline, and their future roadmap to incorporate more acceleration from the GPU Open Analytics Initiative, GOAI.

Photo of Joshua Patterson

Joshua Patterson


Joshua Patterson is a director of AI infrastructure at NVIDIA leading engineering for RAPIDS.AI. Previously, Josh was a White House Presidential Innovation Fellow and worked with leading experts across public sector, private sector, and academia to build a next-generation cyberdefense platform. His current passions are graph analytics, machine learning, and large-scale system design. Josh loves storytelling with data and creating interactive data visualizations. He holds a BA in economics from the University of North Carolina at Chapel Hill and an MA in economics from the University of South Carolina Moore School of Business.

Photo of Michael Balint

Michael Balint


Michael Balint is a senior manager of applied solutions engineering at NVIDIA. Previously, Michael was a White House Presidential Innovation Fellow, where he brought his technical expertise to projects like Vice President Biden’s Cancer Moonshot program and Michael has had the good fortune of applying software engineering and data science to many interesting problems throughout his career, including tailoring genetic algorithms to optimize air traffic, harnessing NLP to summarize product reviews, and automating the detection of melanoma via machine learning. He is a graduate of Cornell and Johns Hopkins University.

Photo of Satish Varma Dandu

Satish Varma Dandu


Satish Varma Dandu is a data science and engineering manager at NVIDIA, where he leads teams that build massive end-to-end big data and deep learning platforms, handling billions of events per day for real-time analytics, data warehousing, and AI platforms using deep learning to improve the user experience for millions of users. Previously, Satish led data engineering teams at startups and large public companies. His areas of interest are in building large-scale engineering platforms, big data engineering, GPU data acceleration, and deep learning. Satish holds an MS in computer science from the University of Houston and is currently enrolled in the management program at Stanford University.