In this paper, we present the architectural details of Odin Help Engine, a novel social search engine that leverages online social networks and sensing data from mobile devices to find targeted answers for subjective queries and recommendation requests. In Odin, users' queries are routed to those connections in their social network who (i) are most likely to help answer the question and (ii) can do so with a high level of confidence. Specifically, we first apply a link-based latent variable model to infer social relationships between users from their social network data to form a strength-weighted relationship graph. We then infer users' expertise by context mining from social network data as well as from their mobile device sensor data. Lastly we apply pagerank-like algorithm that takes both relationship strength and user expertise into account to find a person that is most likely willing to answer the question posted by the user. We present the general design of the architecture and outline a location-related query example for detailed illustration.