To support manifold vertical applications in any deployment environments, the fifth generation (5G) radio access network (RAN) may exploit artificial intelligent (AI) and machine learning (ML) for intelligent RAN configuration. With the Open RAN (O-RAN) architecture supporting the Near-Real-Time (Near-RT) RAN Intelligent Controller (RIC), various AI/ML algorithms can be designed in the form of "xAPPs" to optimize the performance for different vertical applications. To this end, a low-latency Near-RT RIC platform is of crucial importance. The lack of an effective design for the autonomous intelligent xAPP generation for all vertical applications also obstructs the zerotouch operations of the O-RAN. In this article, the new designs to enhance the existing standards and platform of Near-RT RIC, and the new design flow for autonomous intelligent xAPP generation are therefore presented. Experimental results demonstrate the practicability to support various vertical applications.