

Uber's ML Systems (Uber Eats, Customer Support), Declarative Machine Learning - Piero Molino - The Data Scientist Show #064
6 snips Jul 4, 2023
AI Snips
Chapters
Transcript
Episode notes
Graph Neural Networks at Uber Eats
- Piero worked on a graph neural network recommender for Uber Eats that incorporated multiple entities and weighted edges.
- The embeddings reflected changes in user preferences after interactions, improving recommendations noticeably.
Uber's Customer Obsession System
- The "Customer Obsession" ticket system reduced support handling time by 16% without harming customer satisfaction.
- It used multi-task convolutional neural networks combining text data with user behavior and contextual features.
Balance Metrics with Business Goals
- Always balance offline model metrics with online business metrics, especially customer satisfaction.
- Use A/B tests to verify improvements and ensure no negative impact on important business KPIs.