Architecture & ModernizationConference40min
Turning a Distributed Enterprise System into an AI Infrastructure
SPDRM, a global CAE workflow platform, is being adapted into AI-ready infrastructure using MCP and a Gateway pattern. The talk explores how its distributed architecture can coordinate AI agents across enterprise CAE tools while addressing security, orchestration, determinism, and token efficiency.
talk.summaryAiDisclaimer
Yannis KoskinarisCadence (BETA CAE)
Vaggelis SpathasCadence (BETA CAE)
What does it take to evolve a battle-tested distributed system into an AI-ready infrastructure? Can AI assistants meaningfully interact with complex enterprise ecosystems? And how can CAE applications like ANSA, META, and others be centrally coordinated through AI?
SPDRM (Simulation Process Data and Resource Management) is a global CAE platform developed by BETA CAE/Cadence for managing simulation data and workflows, with active deployments across major automotive industries worldwide. Built on WildFly, Spring Boot, and Quarkus microservices — backed by relational databases, full-text search, and a custom distributed file system — it is a large-scale, heterogeneous system shaped by years of real-world engineering demands. Its multi-site architecture connects global engineering teams, centralizing simulation resources while keeping compute close to where it's needed. Its remote job execution engine — handling scheduling, distribution, and lifecycle management across global nodes — proved a natural runtime fabric for agentic workflows.
Evolving this infrastructure toward AI readiness centered on two key architectural choices: MCP and the Gateway pattern, turning SPDRM into a coordination layer for AI agents.
This talk covers the real challenges encountered along the way: enforcing security boundaries, turning CAE applications into MCP servers, orchestrating agentic workflows, balancing deterministic and non-deterministic execution, and minimizing token usage.
SPDRM (Simulation Process Data and Resource Management) is a global CAE platform developed by BETA CAE/Cadence for managing simulation data and workflows, with active deployments across major automotive industries worldwide. Built on WildFly, Spring Boot, and Quarkus microservices — backed by relational databases, full-text search, and a custom distributed file system — it is a large-scale, heterogeneous system shaped by years of real-world engineering demands. Its multi-site architecture connects global engineering teams, centralizing simulation resources while keeping compute close to where it's needed. Its remote job execution engine — handling scheduling, distribution, and lifecycle management across global nodes — proved a natural runtime fabric for agentic workflows.
Evolving this infrastructure toward AI readiness centered on two key architectural choices: MCP and the Gateway pattern, turning SPDRM into a coordination layer for AI agents.
This talk covers the real challenges encountered along the way: enforcing security boundaries, turning CAE applications into MCP servers, orchestrating agentic workflows, balancing deterministic and non-deterministic execution, and minimizing token usage.
Yannis Koskinaris
Yannis Koskinaris is a Senior Software Engineering Manager at BETA CAE Systems (part of Cadence). A lifelong Java engineer with an MSc in Systems Design for Internet Applications (2005), he has worked on SPDRM — BETA CAE's data and simulation process management platform — since its inception, guiding its architecture and evolution over nearly two decades. He leads a team focused on performance, reliability, and developer experience, and is passionate about server-side development, databases, and building systems that scale gracefully as they age.