U.S. citizens donated an estimated 373.25 billion to charity in 2015. These donations came from individuals, corporations, and various foundations. Most of the funds were used for launching projects focused on topics in specific regions of the world. However, there is infrequent formal knowledge transfer between the wide array of projects, and often no singular, unifying historical database. Therefore, an organization initiating a new project may not be aware of what organizations it can partner with, what the estimated value (or the budget) should be, and what learning can be derived from projects that have happened in the past. In this paper, we study publicly available data from the Clinton Global Initiative's Commitment to Action directory, a philanthropic project portfolio comprising 3,200 projects, 10,000 organizations, and multiple topics such as healthcare and education (as of June 2016). We propose a kernel-based tensor factorization approach that provides recommendations to organizations starting on a new project, based on the lessons learned from previous projects.