Graded precision localization is the process by which mobile heterogeneous sensor nodes localize themselves to varying levels of precision, even though they co-exist and operate in a common framework. Some nodes may only need to compute their approximate location at infrequent intervals. Others may need to be continuously and precisely localized. Since GPS is not available indoors, we need to provision a generic ubiquitous localization infrastructure that services specific localization requirements of each mobile sensor node on a best-available-information basis. We assume that each mobile node has an associated user module that gets localized. For implementing graded-precision localization in indoor spaces, we propose a modular architecture for the user module, site-specific deployment strategies, and a framework for evaluating localization performance. The user module architecture is realized using an Android phone and a Cricket sensor node. We report our experimental and simulation results for various operational scenarios.