Explore a 28-minute conference talk that delves into H&M's evolving AI platform, focusing on democratizing and accelerating AI usage across the entire H&M group. Learn about their advancements in speed to production, data abstraction, feature store, and pipeline orchestration. Discover how H&M's reference architecture has been adopted by multiple product teams, managing hundreds of models across the entire value chain. Gain insights into how this architecture enables data scientists to develop models in a highly interactive environment while allowing engineers to manage large-scale model training and serving pipelines with full traceability. Understand H&M's current efforts to reduce time-to-market for new features and shorten the learning feedback loop through AI democratization and adherence to MLOps principles. The talk covers topics such as AI Foundation, Scaling AI, Customer AI, Reference Architecture, ML Code, Release Strategy, Content Personalization, Contextual bandits, Recommendations, and Rewards.
Overview
Syllabus
Intro
Outline
AI Foundation
The Fountainhead
Scaling AI
Fountainhead
Customer AI
Reference Architecture
ML Code
Release Strategy
Content Personalization
Teasers
Slot Containers
Contextual bandits
Recommendations
Rewards
Conclusion
Taught by
Databricks