DoorDash Deploys SafeChat AI Moderation System
DoorDash has built and deployed SafeChat, an AI-driven safety system that moderates text, images, and voice across its platform, handling millions of interactions daily. Using a layered architecture—including an internal model trained on roughly 10 million messages and precise LLMs—SafeChat auto-clears most traffic with sub-300ms responses and 99.8% safe classification, and the company reports about a 50% reduction in low/medium-severity incidents.
Key Points
- 1Implements layered moderation handling text, images, and voice, processing millions of interactions daily.
- 2Trains internal model on roughly 10 million messages to auto-clear most traffic and cut costs.
- 3Enables automated actions, human escalation, and near-real-time responses (sub-300ms) for safety enforcement.
Scoring Rationale
Strong engineering deployment with measurable 50% incident reduction and operational specifics, but limited research novelty beyond system integration and tuning.
Sources
Public references used for this report.
Practice with real Food Delivery data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Food Delivery problems