Unlock the Power of Generative AI
summary
Generative pipeline that combines Amazon Bedrock, SageMaker, Ground Truth, Recognition, Comprehend, Lambda, Fargate and CDK produces image, live-image (parallax) and 6s video assets from a product ASIN with up to four image variants per request and a Cartesian product of prompts × concepts × images. Engineers build a declarative Python SDK and a Lambda orchestrator that calls models via model.invoke so scientists deploy notebooks straight to production. Results: advertisers run 88% more campaigns, see 8% higher CTR, ~177% ROAS lift and scalable online/offline isolation across accounts.
AI-written summary · about
chapters (12)
- 0:00Introduction to generative AI
- 2:12Evolution of generative AI
- 8:13Generative AI in Amazon Ads
- 13:30Image generation architecture
- 18:27Prototyping and model selection
- 23:04Scaling generative AI solutions
- 27:19Responsible AI guardrails
- 31:34Live image generation demo
- 34:03From prototyping to production
- 41:02Declarative workflow development
- 47:13Video generation capabilities
- 53:59Key takeaways and conclusion
related talks
01h 00mSecurely accelerating generative AI innovation
30mGenerative AI in action: from POC to business value
20mBuilding a secure end-to-end generative AI application in the cloud
23mGenerative AI use cases
58mCognizant cognitive architecture for generative AI: The path to MVP
20mExplore what’s possible with new AWS generative AI services
18mNavigating the growth frontiers of generative AI
29mCustomizing genAI applications for your business