Explore a 29-minute conference talk on developing a fully automated and scalable Machine Learning pipeline in the gaming industry. Dive into the experience of an innovative gaming company handling terabytes of data from millions of daily players. Learn how to leverage well-known libraries, frameworks, and efficient tools to create an easy-to-use, maintainable, and integrable ML pipeline. Discover the importance of reproducibility in automated ML pipelines for faster development and easier debugging. Examine how Wildlife uses data to drive product development and deploys data science for core product decisions. Understand the process of improving user acquisition through enhanced LTV models and the use of Apache Spark for distributed computing. Gain insights into the architecture, data processing, model API, training parameters, deployment process, and framework for retraining. Explore additional use cases and learn how Spark enables Data Scientists to run more models in parallel, fostering innovation and onboarding more Machine Learning use cases.
Overview
Syllabus
Introduction
Mobile Gaming
Architecture
Data Processing
Processing at Scale
The Model API
Davinci Library
Dataset
Target Variable
Compound Classifier
Metadata
Training Parameters
Random Forest Classifier
Search Space
Callback
Training
Why it took so long
Run ID
Deployment Process
Deployment to Production
Results
Framework for Retraining
Other Use Cases
Taught by
Databricks
