Worried about your AI IP? Not with ESF!
summary
Edge deployment protects AI by shipping an OpenPGP-encrypted model that ESF decrypts only on-device, streams the decrypted artifacts into a local Nvidia Triton instance, then erases them after loading; rollout uses device snapshots and an Ansible-based batch job from Everyware Software Framework (ESF) and Every Cloud. The demo is a webcam-driven prosthetic hand demo running hand-pose inference on-device, the model encryption/decrypt workflow runs end-to-end, and fleet management, signing, network lockdown and tamper detection are all supported.
AI-written summary · about
chapters (12)
- 0:00Protecting AI models at edge
- 1:08Deploying AI on edge devices
- 3:39Edge AI for real-time analytics
- 5:04IoT application challenges
- 6:50ESF software framework
- 12:11Operational workflow with ESF
- 15:55Industry applications using ESF
- 16:40Security risks in edge AI
- 19:50Integrated security for edge devices
- 22:07Protecting AI models with encryption
- 23:16Live demo: AI model protection
- 26:50Conclusion & Q&A
related talks
38mStreamlining AI Solutions Development: S...
22mEdge AI Use Cases in IoT — Unlocking Intelligence at the Network Edge
55mEdgeAI - AI in the Palm of Your Hand
20mDeploying AI in Days, Not Months: Siemens Industrial Edge on
20mScaling Edge AI with Elixir and Nerves
28mGenerative AI + Proprietary Data
19mHow to Deploy Artificial Intelligence in Industrial Application
36mApplied AI on the edge