Secure Agentic Workflows in Finance: V. Caldeira & M. Foster Demonstrate Kubernetes Reference Architecture on PyTorch Channel

V. Caldeira and M. Foster released a talk titled "Architecting Secure Agentic Workflows on Kubernetes: A Financial Sector..." on the official PyTorch channel on September 24, 2026. The resource addresses the infrastructure complexities organizations face when moving from basic LLM integrations to autonomous agentic workflows, specifically highlighting challenges in tool discovery, secure access, and traffic routing for multi-agent architectures in production.

The speakers explore leveraging Kubernetes and cloud-native abstractions to develop, test, and deploy AI agents at scale. Central to the presentation is a reference architecture based on the Kagenti project, designed to construct a secure and scalable agentic environment. Key technical components include unifying tool access via an MCP Gateway, enforcing zero-trust workload identity with SPIFFE/SPIRE, and standardizing inter-agent communication.

To illustrate these concepts, the talk walks through a real-world financial use case: an autonomous agent that securely accesses market data and executes simulated transactions using financial tools over MCP. This demonstration highlights end-to-end cloud-native deployment, providing a concrete architectural reference for engineering teams focused on AI infrastructure security.