This study presents a framework of platform-based inference system design to provide specific field experts/users developing customized fuzzy inference system efficiently and effectively. The framework is composed of three major components: FML editor, FML parser and FML inference engine. First, the FML editor can let field experts compile knowledge base and rule base with friendly user interface. The composed FML documents are then sent to our inference system platform through the Internet. The FML parser is used to transform the predefined knowledge and rule bases into object-oriented database. The inferred results eventually can be derived by FML inference engine with testing data collected by a user and the interpreted object-oriented database. Besides, users also can directly use our inference system platform if related FML-based knowledge and rule bases defined by other experts. The convenience and practicability for developing a fuzzy expert system in a specific field have been demonstrated by its application to health-care and abnormal diagnosis of cogen systems.