Respectful human-robot interaction is informed by local culture. This chapter describes an ongoing project to equip Pepper humanoid social robots with the ability to interact verbally and non-verbally with people in Rwanda and South Africa in a culturally sensitive manner. We describe an exercise to acquire Rwandan cultural knowledge by means of an online survey, targeting spatial, verbal, and non-verbal communication. Each type of knowledge in the survey is structured using a cultural knowledge ontology. We describe the development of a complete system architecture that uses this cultural knowledge to interact with people, illustrating the approach with one specific example in which a Pepper robot gives visitors a short tour of a robotics laboratory. The system is implemented in the Robot Operating System (ROS) by developing a package of ROS nodes that provide functionality for overt attention (based on detection and localization of sounds, people, and faces), speech understanding and production, animate behavior, deictic, iconic, and symbolic gesture, and navigation. The robot mission is coordinated by a ROS behavior controller node that interprets the use case scenario specified by a behavior tree.