«Hands-on Image Recognition with Scala, Spark and DeepLearning4j», Guglielmo Iozzia
Who said that Python is the only programming language choice for Deep Learning matters? This talk will give first a 101 intro to DeepLearning4J, an Open Source distributed framework integrated with Hadoop and Spark that can be used on CPUs and GPUs and has, despite the final “J” in its name, support for Scala. Then it will walk through some simple image recognition example based on DL4J and Spark.
«Tracking with Writer Monad», Eugene Platonov
This talk will tell the story of one team at eBay which used to do data tracking in a healthy side-effecting manner. Until the Team realized that it’s not that healthy. Solution was found in a Writer Monad (residing in cats library) as well as in the fact that Writer Monad can stay in shades. Some people, especially whey they are new to typed FP, don’t like/feel comfortable to see words like Semigroup, Traversable, Writer and such in their domain code. The talk will show how those “scary” parts can be “hidden” by domain specific extension methods.
«User Behavior Analytics for Cyber Security and Its Implementation In Scala», Derek Lin & Leonid Kladko
More than a promise, User Behavior Analytics (UBA) is here and now in identifying malicious activities in enterprise networks. We walk you through a UBA system grounded on data science and enabled by an architecture implemented in Scala and open source technologies. In particular, we introduce one machine learning use case for abnormal behavior detection and do a deep dive on its Apache-Spark based implementation.
«Evolution of ads management platform architecture from Akka Cluster to Akka Streams Kafka», Andrew Lashchenko
During this presentation, we will carefully examine the evolution of the ads management platform’s architecture that has been an existing process for over 3 years. The original intent was to generate recommendations based on analysis of reporting data that was obtained from various programmatic advertising platforms. Thus, the system was created exclusively for the internal use by a team of campaign managers. This project turned into a complete platform with full capabilities to create and manage advertising campaigns across premium digital destinations and target over 25 premium native, social and content discovery channels. This presentation will outline how ads management platform’s architecture has changed over time, why the original decision was to use the Akka Cluster, numerous issues and challenges encountered during this project with their respective solutions and transition from the batches to streaming processing that utilizes Akka Streams Kafka.
«Practical Introduction to Basic Cats Typeclasses», Anatolii Kmetiuk
Purely functional programming is a powerful style that allows to reduce mental load on the programmer and make the code more modular and less interdependent. To facilitate programming in functional style, libraries like Cats were created. In this talk, we will discuss the motivation for and usage of the basic type classes of Cats library.