As artificial intelligence reshapes the future of mobility, autonomous driving is emerging as one of the most demanding and safety-critical frontiers for AI innovation. From perception and decision-making to large-scale simulation, validation and cloud-native development, the journey toward safe autonomy increasingly depends on advanced AI agents, foundation models and scalable computing infrastructure.

In this interview with AI Spectrum, Yann Baudouin, Head of Data Driven Development and DevOps in AUMOVIO’s business area Autonomous and Commercial Mobility, shares how agentic and generative AI are transforming the way autonomous systems are developed, tested and validated. He explains how AUMOVIO’s collaboration with AWS is accelerating simulation and safety workflows, why freight autonomy is emerging as a compelling first use case, and how cloud-native AI is becoming essential to building adaptive and reliable autonomous vehicles.

Baudouin also offers a forward-looking perspective on how agentic AI and foundation models could reshape regulation, safety frameworks and public trust over the next five years, outlining the path toward more robust, scalable and trusted autonomous mobility systems.

How do agentic and generative AI agents change the development and testing of perception and decision-making systems compared to traditional simulation?

Using agentic and generative AI agents is fundamentally changing our development and testing activities. Moving on from static, real world simulations we are now able to create adaptive environments where our products in the development phase are faced with unpredictable and more borderline challenges. This enables higher realism, continuous learning and scalable testing, ultimately leading to improved robustness and faster development times.

How does combining AUMOVIO’s automotive expertise with AWS’s infrastructure help compress validation timelines while maintaining safety standards?

Combining AUMOVIO’s automotive expertise with AWS’s cloud infrastructure accelerates validation by enabling massively parallel, high-fidelity simulations at scale with maximal service availability at the same time. This approach maintains safety standards through automated compliance checks, secure data handling, and continuous integration pipelines, ensuring rigorous testing without sacrificing speed.

How does this partnership improve the speed and accuracy of identifying critical safety patterns and edge cases from real-world driving data?

Developing autonomous vehicles requires processing vast amounts of data from millions of road scenarios and safety conditions. By combining AUMOVIO’s automotive expertise with AWS’s scalable cloud and AI capabilities, including Amazon Bedrock, the collaboration enables rapid, high-volume data processing. The use of generative AI technologies reduces search times to a great extent. This accelerates the identification of critical safety patterns and significantly shortens the timeline for improving hazard detection.  

Regarding the deployment of autonomous trucks for Aurora, what makes freight a compelling first use case, and what can be applied to broader mobility?

Freight is a compelling first use case for autonomous trucks in a hub-to-hub scenario because it involves more predictable highway routes, fewer complex urban interactions, and high economic impact from efficiency gains, making it easier to scale and deliver immediate value. Things learned here, can be then at a later time also transferred towards the passenger car market.

As AWS is now your preferred provider, how essential is cloud-native AI to the future of autonomy?

Cloud-native AI is essential to the future of autonomy because it provides scalable, on-demand compute and storage, enabling real-time processing of massive sensor data and simulations, while supporting continuous learning and deployment. This architecture drives faster innovation, improves cost efficiency, and enables seamless integration of advanced AI models—key factors for building safe and adaptive autonomous systems.

How do you see agentic AI and foundation models reshaping the autonomous vehicle ecosystem over the next five years regarding regulation and safety?

Over the next five years, agentic AI and foundation models will fundamentally reshape the autonomous vehicle landscape, particularly concerning regulation and safety. We anticipate a significant shift towards 'Safety for AI,' where AI systems themselves are rigorously evaluated and certified for safety-critical applications. Agentic workflows, combined with deep data analytics, will unlock unprecedented capabilities for searching and reporting on vast datasets, enabling faster identification and mitigation of potential safety risks. This will lead to more comprehensive scenario coverage for safety case development, ensuring autonomous vehicles are prepared for a wider range of real-world driving conditions. Ultimately, these advancements will pave the way for more robust regulatory frameworks and increased public trust in the safety of autonomous driving technology.