MLOps.community Inside Uber’s AI Revolution - Everything about how they use AI/ML
92 snips
Jul 4, 2025 Kai Wang, product lead of Uber's AI platform team, dives into the fascinating world of Uber's Michelangelo, the internal ML platform that powers critical business use cases. He discusses how Uber's approach to AI goes beyond buzzwords, focusing on tools for both novice and advanced users. Kai sheds light on predictive and generative models, detailing how they enhance services like ride booking and Uber Eats. He also hints at exciting plans to open-source parts of their framework, underscoring the importance of community collaboration in AI innovation.
AI Snips
Chapters
Transcript
Episode notes
Broad Spectrum of AI at Uber
- AI at Uber goes beyond just large language models like ChatGPT, including simpler models like linear regression and random forests.
- This broad spectrum of AI techniques is integrated to serve various business needs effectively.
Michelangelo's Creation Story
- Early ML usage at Uber started with rule-based replacements but involved chaotic one-off code in notebooks.
- Michelangelo was created in 2016 to provide a centralized, end-to-end machine learning platform to scale and standardize projects.
Empower Developers with Choices
- Let developers choose their tools: provide easy options for most users and direct infra access for power users.
- Use orchestration frameworks like Uniflow to allow customization and flexibility in machine learning workflows.