The Internet affords new approaches to learning. Geographically dispersed self-directed learners can learn in computer-supported communities, forming social learning networks. However, self-directed learners can suffer from a lack of continuous motivation. And surprisingly, social learning networks do not readily support effective, coherence-creating and motivating learning settings. It is argued that providing project-based learning opportunities and team formation services can help overcome these shortcomings. A review of existing team formation tools evidences that a new design for team formation and the initiation of project-based learning is required before these can be supported in social learning networks. A design is proposed which identifies "knowledge", "personality" and "preferences" as categories in which data is needed to form teams, and it specifies how the required data are gathered and assessed. The design defines rules deduced from team formation principles from prior team formation research to optimise team formations towards increased productivity, creative solutions or higher learning outcomes. The rules are implemented in three team formation expressions each calculating one of the desired team formations. The expressions are deployed on a set of test data, demonstrating the effectiveness of the team formation service design. The article includes a discussion of the results and provides indications for future research.