Dive into a 49-minute podcast episode featuring Jean-Michel Daignan, a data scientist at Ubisoft, as he shares insights on machine learning in production from a data scientist's perspective. Explore the unique challenges and solutions in developing platforms and tools for data scientists in the gaming industry. Learn about Ubisoft's approach to tying machine learning products to business use cases and ROI, as well as their use of generative AI. Gain valuable knowledge on balancing quick delivery and value creation, end-to-end pipelines, data quality ownership, monitoring systems, and feature stores. Discover how Ubisoft leverages play data to understand player personalities and implements deep learning models. Get a glimpse into the massive data volumes handled by Ubisoft and the innovative ways they service data scientists' needs in a fast-paced gaming environment.
Machine Learning in Production: A Data Scientist's Perspective from Ubisoft
Overview
Syllabus
[] Jean-Michel's preferred beverage
[] Jean-Michel Daignan's background
[] Takeaways
[] Rate us and share the podcasts with your friends!
[] Jean-Michel's projects at Ubisoft
[] Jean-Michel's success as a Data Scientist
[] Ubisoft basics
[] Jean-Michel's success from the downfalls of being a data scientist
[] Building for data scientists' considerations
[] Differences in designing for data scientists in general
[] End twin pipelines and their functions
[] Major problems doing maintenance
[] Data quality ownership
[] Monitoring levels
[] Locomotive systems
[] Merlin
[] DS storage systems
[] Feature stores batch or streaming?
[] Bringing Machine Learning to Production at Ubisoft blog post
[] Features and recommendation systems
[] Playing games
[] Play data = play personalities
[] Deep learning in all the diffusion models or the foundation models
[] Servicing data scientists' needs
[] Ubisoft's data volume
[] Wrap up
Taught by
MLOps.community
