Architecting Secure Agentic Workflows on Kubernetes: A Financial Sector...
Talkat KubeCon + CloudNativeCon + OpenInfra Summit + PyTorch Conference China 2026 · 2026-09-07..09-09·
summary
Kubernetes-based reference architecture uses Rosoctl CRD + sidecar to inject short‑lived SPIFFE identities, Keycloak JWTs and route tool calls through an MCP Gateway; demo deploys a financial agent that fetches market data and executes simulated transactions via MCP servers, then triggers a malicious portfolio-audit injection that observability (MLflow traces) and an LLM judge catch and block in real time. Platform runs 100,000+ simulated scenarios for evaluation-driven development and samples 10% of traces for automated auditing.
AI-written summary · about
chapters (14)
- 0:00Introduction
- 1:04Agentic workflows foundations
- 3:11Cloud Native Agentic Standards
- 7:27Zero trust for agents
- 9:14Rosoctl and defense in depth
- 13:50Namespace isolation for agents
- 14:57Agent as custom resource
- 18:30Platform security shift
- 20:37Evaluation-driven agent dev
- 25:12Live demo setup
- 29:49Agent injection attack
- 32:18Fixing agents with LLM judges
- 35:55Final architecture benefits
- 36:42Synthesis and future
related talks
31mArchitecting Secure Agentic Workflows on Kubernetes: A Financial Sector…
27mKeeping Your Agents in Check: Layered Security for Agentic Platforms in Production
37mAchieving High Security at Scale with Microservices
25mWho Let the Agents Out? Securing AI Workflows the Right Way
25mMulti-messenger Security: Adaptive Kubernetes SOC...
57mThe Kubernetes Sessions Security
36mSecuring Agentic Platforms with Zero-Trust Workload Identities
20mSimplifying Kubernetes: Streamlining Secure App Deployment in DevSecOps