Explore the challenges and solutions for productionizing real-time machine learning models in enterprise environments in this 29-minute conference talk. Dive into the complexities of ML data engineering and learn how to accelerate the delivery of ML applications using a feature platform. Discover how Spark and DataBricks provide a scalable foundation for data engineering, and understand how a feature platform extends data infrastructure to support ML-specific requirements. Gain insights from Atlassian's successful deployment of a feature platform for real-time, ML-driven personalization and search services. Learn about managing feature transform logic, generating high-quality training sets, and implementing end-to-end feature lifecycle management. Explore a practical example of automated content categorization in Jira to see these concepts in action.
Overview
Syllabus
Intro
Building Operational ML applications is very complex Data is at the core of that complexity.
Features are the signals we extract from data and are a critical part of any ML application.
Tecton is a data platform for ML applications
Managing sprawling and disconnected feature transform logic
Building high-quality training sets from messy data
Configuration-based training data set generation through simple APIs
1 Configure what features you want in a training dataset
Built-in row-level time travel for accurate training data
End-to-End Feature Lifecycle Management
Example: Automated Content Categorization in Jira
Taught by
Databricks
Related Courses
-
Data Engineer, Big Data and ML on Google Cloud auf Deutsch
-
Spark, Hadoop, and Snowflake for Data Engineering
-
MLOps Platforms: Amazon SageMaker and Azure ML
-
AI at Scale: Industrializing ML with Databricks
-
Hopsworks - The Python-Centric Enterprise Feature Store
-
Enable Production ML with Databricks Feature Store
