As software defined car becomes a reality, various censoring technology, AI algorithms have become a normal for a vehicle which is expected to be equipped with different level of intelligence. We focus on the computing platform which is used to implement such technology and argue the only way to achieve cost-effective, power-efficient autonomous driving system is to embrace heterogeneous computing architecture we are now seeing from various silicon implementation. Instead of close-door engineering, we demonstrate through open collaboration and collective engineering across different tiers in the ecosystem, we are able to accelerate the development of technology required for autonomous system. We show case how open standard can enable silicon vendors, Tier 1 system integrators and OEMs fast prototyping their PoC projects with a clear path for production. We also argue standard interface is required and beneficial for various heterogeneous computing platforms to best protect the investment at the application level of such systems. This is particular true for generic features such as safety, redundancy deployed in various E/E architecture. Through 96Boards Automotive, Autoware Foundation and their partners, we demonstrate how a real world autonomous driving system is built using heterogeneous SoCs, low cost and high power efficient system design with open source full stack software equipped with clear path to functional safety production system.