This paper describes an agent-based recommendation system developed to support knowledge acquisition and sharing processes. Its purpose is to aid the process of community building, specially in the early phase of peer search and recognition in a distributed engineering environment, where multidisciplinary teams must be established. In this kind of problem information about possible candidates to make up the teams must be gathered from all available channels, including the electronic ones. Due to privacy issues, web based data mining systems are not efficient for the purpose of constructing candidate?s profiles, even if they can offer some help; therefore a distributed solution that permits each candidate?s workplace to be explored was developed. As a means of aggregating collected knowledge for sharing, a conceptual model, relating candidate?s profiles and their knowledge domain was developed in the form of an ontology. This strategy allows the system to be specialised in terms of the candidates? main knowledge domain.