Considerable progress has been made in recent years in locating and recognizing faces using advanced machine learning and computer vision techniques and a number of interactive tools are available for general use by individuals to use these technique in an ad hoc manner. However, no known, easily accessible open source framework presently exists to meet organizational needs to search for given faces against their pre-stored image sets, either in data analytics efforts or in time-critical situations, which leverages these techniques. Consequently, at the US National Library of Medicine (NLM), we have implemented a Web based, publicly accessible system called FaceMatch (FM), which provides customized face matching services to clients through programmatic interfaces to robust face recognition software. In addition, it provides tools for use by the clients to submit requests in interactive or batch environment and visually observe the returned results. Two key aspects of the FaceMatch system are: (a) it stores the contents of client's images in a repository through their key features instead of pixel values, avoiding potential copyright and other legal problems; (b) it assures near real-time availability of a newly ingested image for subsequent searches, eliminating perceptible delay between ingest and query, which is quite important in timecritical situation such as a natural or manmade disaster. In addition to providing an HTML/REST-based API for clients to send requests to the FaceMatch server programmatically, the system also provides tools (integrated into a Java application called the FM Workbench), which allow users to submit requests interactively or in batch mode, and to visualize the returned results in real-time. The FaceMatch system has been built for use with NLM's PEOPLE LOCATOR® Service, replacing an earlier, proprietary visual search system, and is available to be used by others with similar needs.