AI at Amazon
summary
Amazon reports 4,500 developer years saved (about $260M) with developer tooling; Q CLI (Claude backend) generates 80% production-acceptable code and can design, test and deploy apps. Merchant-facing Project Amelia provides conversational workflows and generative product imagery used by millions. Alexa upgrades combine multimodal voice models and Nova agents for natural task handling. Warehouses run ~750,000 robots and new tactile Vulcan robots use synthetic data to learn fragile handling. Nova Act and Strands SDKs enable browser-operating and long-running agentic apps.
AI-written summary · about
chapters (12)
- 0:00Introduction to AI at Amazon
- 0:51Amazon's AI product offerings
- 2:18AI impact on developer productivity
- 3:42Project Amelia for merchants
- 5:30Generative AI for product imagery
- 6:34Alexa's generative AI redesign
- 8:45Robots in Amazon warehouses
- 10:52Vulcan robots with tactile sensation
- 11:56Synthetic data for robot training
- 12:17AI agents operating complex tools
- 13:20Amazon's agent SDKs
- 14:24Optimizing AI model performance
related talks
06mAI and machine learning at Amazon
18mHow Amazon builds and deploys autonomous robots
30mGenerative AI in action: from POC to business value
46mThe AI-Powered Developer
21mGen AI disrupts SDLC. What does it mean for developers? The AWS approach | AWS Developer Day 2025
18mAI Unleashed: Forging the Future of Rapid Innovation and Growth
27mSoftware engineering with GenAI
32mBuilding a GenAI-Driven Developer Organization