Traditional scientific computing systems relied on consecrated network of systems that arrives with complex maintenance and are deficient in flexibility and scalability. These issues necessitate immense level configuration process in result of a change in the setup (changing of resources, change of platform). Cloud computing, an on demand and cost effective means of provisioning computing resources, leverages all the above difficulties. The practical applications of clouds for scientific computing applications are still one of the largest unexplored areas in the cloud computing domain. Java Enterprise Platform (J2EE) has a successful journey in the field of distributed interoperable systems. This study explores the J2EE capabilities such as dynamic service discovery based on publishing of XML based configuration information of the service to the cloud. The Optimal Cloud Platform Selection (OCPS) algorithm evaluates the clouds for the optimal deployment of the scientific computing services based on various computational parameters like ram, processor cycles etc along with computational cost under consideration. This framework is flexible enough to cope with the dynamic nature of the clouds. The implementation has been tested in a hybrid cloud infrastructure using eucalyptus open source private cloud platform along with Google app-engine as a gateway to the outside world.