Defense against the dark arts of AI: The playbook for a sovereign model-as-a-service platform
summary
Agenda shows a practical playbook: build a sovereign MaaS by combining GPUs + inference servers (LMD/Llama-style), a distributed inference layer, an API gateway and service mesh, plus Backstage scaffolding templates that instantiate a RAG chatbot in ~5 minutes. Expect 95% of pilots fail without platform engineering; use vector DBs for RAG, Granite/HuggingFace models, Quarkus apps, Argo rollouts for mirroring and canaries, TrustyAI for input/output guardrails, and monitoring + cost metrics to detect surprises.
AI-written summary · about
chapters (15)
- 0:00Model-as-a-service and AI sovereignty
- 3:31Introduction
- 6:44Geopolitical risks and data sovereignty
- 9:11Sovereign AI platforms metaphor
- 11:16Technical stack of model-as-a-service
- 15:30AI inference analogy
- 17:17Model selection and cost optimization
- 19:05RAG vs fine-tuning
- 22:18Agentic flows and autonomous AI
- 24:50Platform engineering demo
- 27:42Service mesh and deployment patterns
- 31:39Secure supply chain for AI models
- 34:11Guardrails and protection demos
- 39:07Recap and building sovereign AI
- 40:36Resources and closing
related talks
27mThe playbook for a sovereign Model-as-a-Service platform
34mUnlock the full potential of Generative AI with Agent Platforms
30mBeyond Monolithic AI: Cloud Native Patterns for Dynami...
38mBuilding AI responsibly with a GRC strategy
01h 01mSoftware Architecture for Gen AI
46mIntelligence as Infrastructure: What Developers Must Learn to Build with AI
30mAI Application Architecture: Building for Scale and Efficiency
29mOpen Source Sovereign AI: Building Enterprise GenAI Platforms for Control & Comp...