Over the past decade, autonomous driving development has undergone a profound transformation—from fragmented ECU-based architectures and smart sensors to centralized, software-defined vehicle platforms powered by AI. Drawing on experience across academia, startups, and enterprise environments, the talk will explore how system architectures, sensor data & perception pipelines, and development workflows have evolved in production vehicles. We’ll also examine how these shifts have reshaped R&D practices, tooling, and team competencies, both in terms of hard and soft skills, organization structure, and modus operandi. As AI continues to expand beyond perception into planning, control, and end-to-end systems, we’ll take an outlook to what this means for future hardware/software co-design and engineering organizations.

• The architectural shift:from distributed ECUs to centralized, high-performance SDV platforms
• Evolution of perception pipelines: from modular pipelines to integrated, data-centric systems
• Changing R&D workflows: simulation, CI/CD, and data-driven development at scale
• New skill sets and tooling: what modern AD teams need vs. a decade ago
• The future: AI beyond perception, end-to-end approaches, and HW/SW co-optimization