The paper is devoted to knowledge representation technologies, reasoning models and algorithms for generating cognitive hypotheses in intelligent systems. The emphasis is placed on applying problem—oriented graphs of knowl­edge in nuclear physics and nuclear power. The proposed technologies and methods cover the tasks of computer— aided detection and classification of nuclear knowledge and competences based on ontologies and representation of information objects in semantic graph data bases equipped with automated reasoning means. Templates for knowledge graphs designing for nuclear research and training centers are presented. Visual navigation and reasoning on the knowledge graphs are performed by means of special retrieval widgets and the smart RDF browser. The visual way of specifying the inference rules on the knowledge graphs is the highlight of the reasoner in the proposed project, which makes it stand out from the more traditional known reasoners, where inference rules are specified using logical predicates and SPARQL—like syntax. Operations with the semantic repository are implemented on cloud platforms through SPARQL queries and RESTful services. The proposed software solutions are based on the cloud comput­ing using DaaS, IaaS and PaaS service models to provide scalability of data warehouses and network services. The architecture of the software in UML notation is presented, examples of real use of the created technologies and software are given.