Image mode for AI inference across the hybrid cloud
summary
Demonstration builds a RHEL bootable image by containerizing a Ramal Llama model and chatbot with Podman, then using the bootsy image builder to produce a raw disk, snapshot to S3, and convert to an AWS AMI that boots an EC2 instance and runs the chatbot successfully. Build environment uses Fedora cloud base, cloud-init, IAM/SSM and private registries or Quay; update strategy is tear down and rebuild, with S3 snapshot swaps for fastest model rotation.
AI-written summary · about
chapters (7)
- 0:00Image mode and AI in cloud
- 6:52Bootable containers for AI infra
- 11:28Simple tools for AI infra
- 13:18Live demo: AI model with image mode
- 21:31Isolated AI dev environments
- 27:34AI workloads in hybrid cloud
- 28:17Q&A: update strategies
related talks
52mPublic Cloud Meetup
25mCreate your own Fedora image & push it to cloud
35mAccelerate deep learning in the cloud with custom ML environments
19mAGL in the Cloud: Progress on AMI Builds
01h 14mModernizing and Optimizing Applications for the Cloud
38mUp and running with Fedora CoreOS
35mBuilding Enterprise Solutions in the AI-Native Era
21mAWS On Air ft. Amazon SageMaker Inference Recommender