Sustainability architecture design for ML workloads
summary
Outlines a shared responsibility model where AWS covers data-center efficiency and renewables while customers optimize code, data lifecycles and instance choices; cites a 451 Research finding of 3.6× energy efficiency and 88% lower carbon footprint versus median enterprise data centers; recommends using managed offerings like Amazon SageMaker (checkpointing, auto shutdown), SageMaker Neo, S3, purpose-built silicon and elastic inference, automating build/endpoint scaling and retraining only on measured model drift.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 1:16Agenda
- 2:26Sustainability of AWS
- 4:54Sustainability in the Cloud
- 12:23Instance Selection
- 14:41Deployment Management
- 16:47Sustainability through the cloud
- 18:32Conclusion
related talks
32mOptimizing AI/ML workloads for sustainability
25mDesigning an AWS Well Architected Framework for AI/ML
20mBuilding GenAI That Doesn't Cost the Earth
48mAccelerate science, research, and innovation with AWS
20mThe 3 energy saving tips for your data science ecosystem
34mLat Designing and Building Complex Machine Learning....
12mLeveraging AWS and Kubernetes for Scalable AI/ML Workloads in the Cloud
34mDeploying a Machine Learning System That Only Runs Locally