Schedule: Media, Marketing, Advertising sessions

Abhishek Kumar and Pramod Singh walk you through deep learning-based recommender and personalization systems they've built for clients. Join in to learn how to use TensorFlow Serving and MLflow for end-to-end productionalization, including model serving, Dockerization, reproducibility, and experimentation, and Kubernetes for deployment and orchestration of ML-based microarchitectures. Read more.

As a fully closed model economy, games offer a unique opportunity to use analytics to create unique purchase opportunities for customers. Bysshe Easton and Thomas Dobbs explain how KIXEYE uses machine learning to create personalized offer recommendations for its customers, resulting in significantly increased monetization and retention. Read more.

Hundreds of thousands of ETL pipelines ingest over a trillion events daily to populate millions of data tables downstream at Netflix. Jitender Aswani, Girish Lingappa, and Di Lin discuss Netflix’s internal data lineage service, which was essential for enhancing platform’s reliability, increasing trust in data, and improving data infrastructure efficiency. Read more.

The Netflix data platform is a massive-scale, cloud-only suite of tools and technologies. It includes big data tech (Spark and Flink), enabling services (federated metadata management), and machine learning support. But with power comes complexity. Kurt Brown explains how Netflix is working toward an easier, "self-service" data platform without sacrificing any enabling capabilities. Read more.

Customer segmentation based on coarse survey data is a staple of traditional market research. Melinda Han Williams explains how Dstillery uses neural networks to model the digital pathways of 100M consumers and uses the resulting embedding space to cluster customer populations into fine-grained behavioral segments and inform smarter consumer insights—in the process, creating a map of the internet. Read more.

Google uses deep learning extensively in new and existing products. Join Ron Bodkin to learn how Google has used deep learning for recommendations at YouTube, in the Play store, and for customers in Google Cloud. You'll explore the role of embeddings, recurrent networks, contextual variables, and wide and deep learning and discover how to do candidate generation and ranking with deep learning. Read more.

Eric Bradlow and Zachery Anderson discuss the Wharton Customer Analytics Initiative research opportunity process and explain how some of EA’s solved some of its business problems by sharing its data with 11 teams of researchers from around the world. Read more.

Boris Yakubchik and Salah Zalatimo offer an overview of Bertie, Forbes's new publishing platform—an AI assistant that learns from writers and suggests improvements—and detail Bertie’s features, architecture, and ultimate goals, paying special attention to how the company implemented an ensemble of machine learning models that, together, make up the AI assistant's skill set and personality. Read more.

Using a messaging system to build an event bus is very common. However, certain use cases demand a messaging system with a certain set of features. Sijie Guo and Penghui Li discuss the event bus requirements for Zhaopin.com, one of China's biggest online recruitment services providers, and explain why the company chose Apache Pulsar. Read more.

Video is now the largest source of data on the internet, so we need tools to make it easier to process and analyze. Alex Poms and Will Crichton offer an overview of Scanner, the first open source distributed system for building large-scale video processing applications, and explore real-world use cases. Read more.

Today, normal growth isn't enough—you need hockey-stick levels of growth. Sales and marketing orgs are looking to AI to "growth hack" their way to new markets and segments. Ken Johnston and Ankit Srivastava explain how to use mutual information at scale across massive data sources to help filter out noise and share critical insights with new cohort of users, businesses, and networks. Read more.

Netflix has over 125 million members spread across 191 countries. Each day its members interact with its client applications on 250 million+ devices under highly variable network conditions. These interactions result in over 200 billion daily data points. Vivek Pasari dives into the data engineering and architecture that enables application performance measurement at this scale. Read more.

Data has become a foundational pillar for security teams operating in organizations of all shapes and sizes. This new norm has created a need for platforms that enable engineers to harness data for various security purposes. John Bennett and Siamac Mirzaie offer an overview of Netflix's internal platform for quickly deploying data-based detection capabilities in the corporate environment. Read more.

Brands that test the content of ads before they are shown to an audience can avoid spending resources on the 11% of ads that cause backlash. Using a survey experiment to choose the best ad typically improves effectiveness of marketing campaigns by 13% on average, and up to 37% for particular demographics. Patrick Miller explores data collection and statistical methods for analysis and reporting. Read more.

The ICIJ was the team behind the Panama Papers and Paradise Papers. Pierre Romera offers a behind-the-scenes look into the ICIJ's process and explores the challenges in handling 1.4 TB of data (in many different formats)—and making it available securely to journalists all over the world. Read more.

With so much data being generated in real time, what if we could combine all these high-volume data streams and provide near real-time feedback for model training, improving personalization and recommendations and taking the customer experience to a whole new level. Sonali Sharma and Shriya Arora explain how to do exactly that, using Flink's keyed state. Read more.

Decision making often struggles with the exploration-exploitation dilemma. Multi-armed bandits (MAB) are a popular reinforcement learning solution, but increasing the number of decision criteria leads to an exponential blowup in complexity, and observational delays don’t allow for optimal performance. Shradha Agrawal offers an overview of MABs and explains how to overcome the above challenges. Read more.