Most AI agent demos skip what matters in automotive engineering: where the agent's code runs, what it can reach, and how anyone knows its answers are right. We present an architecture that uses a managed agent service for scale while keeping agent sandboxes inside our own cloud. We also introduce an agent development workflow that integrates prompt, skills, tools, and evaluation assets. This workflow enables agent context that is reproducible at the byte level, release-gated on evaluation scores. This is not an agent productivity pitch; it is a walkthrough of the architecture that makes agentic work reproducible and measurable.

The commercial vehicle industry is undergoing one of the most significant technology transitions in its history. As a highly cost-sensitive market segment, commercial vehicles are among the first to adopt software architectures designed to reduce hardware, integration, and lifecycle costs. Regulatory requirements for cybersecurity and software lifecycle management, combined with the migration toward centralized computing architectures and Software-Defined Vehicles (SDVs), are fundamentally changing how software is developed, deployed, and maintained across vehicle platforms. Traditional integration models that tightly couple software functions to proprietary hardware are rapidly becoming barriers to innovation, scalability, and cost-effective deployment.

Together, ETAS and Cummins are demonstrating a new approach based on software portability. Rather than adding another telematics unit, supplier applications can run directly on existing OEM computing infrastructure through a combination of industry standards and open-source technologies.

This enables a "build once, deploy anywhere" model for vehicle software.

The solution combines ASAM/ISO SOVD for standardized vehicle access with Eclipse SDV technologies including S-CORE, Symphony, uProtocol, and KUKSA for orchestration, communication, and data management. Eclipse VAMUS (Vehicle Adaptive Motion Unification Service) extends this approach by demonstrating how portable software services can coordinate vehicle motion functions across propulsion, transmission, and braking systems. This creates a common software foundation while allowing Cummins to retain ownership of its proprietary applications and intellectual property.

The outcome is reduced hardware redundancy, lower integration costs, faster software deployment, and greater scalability across OEM platforms. More broadly, the project demonstrates how standards can provide interoperability while open-source collaboration accelerates innovation. Together, ETAS and Cummins are establishing a practical blueprint for the future software-defined commercial vehicle ecosystem, where shared infrastructure enables innovation and proprietary software delivers competitive differentiation.

The transition to software-defined vehicles (SDVs) is no longer a future vision - it is a present-day operational imperative for every major automotive manufacturer. Yet , significant questions remain: How are OEMs making critical SDV decisions? Which software architectures are emerging as industry standards? What capabilities are they actively seeking from suppliers? And where does the gap between ambition and execution still exist? 

In this session, Prashant Gulati CEO of SDVerse provides a perspective on how the world's leading automakers are navigating the SDV transformation. Drawing on public announcements and statements from General Motors, Ford, Stellantis, and other major OEMs across North America, Europe, and Asia, Prashant shares real-world insights into the priorities, challenges, and opportunities shaping the future of automotive software. 

Moving beyond industry theory, this presentation examines how OEM leaders are structuring software architecture strategies, evaluating sourcing decisions, and adapting procurement models to meet the increasing complexity of software-driven vehicle development.