ScarabResearch is a startup company providing recommendation engines and data mining services to on-line retailers. While new in the business, Scarab is the engine behind Hungary's largest on-line store, just won a contract for one of the largest on-line food stores in the US and the currently launching Amazon-EC2 / Google AppEngine cloud solution will serve subscription customers. In our year or so long operational experience, we have seen a very versatile palette of data sources (ranging from bookstores, grocery stores to auction sites) that provided challenging problems as well as valuable business insights. The talk will focus on the architecture of our recommender system, the algorithms we currently use and research and on vendor and domain specific insights that were factored into the algorithms to provide better recommendations.