The team developed a generalized personalization system for World-Wide-Web commerce sites. The system provides a framework for giving the users of a commerce site a personalized buying experience, such as recommending products specifically chosen to match the user's interests. The team developed and implemented several algorithms to filter user information and recommend products, by employing techniques from the areas of neural networks and collaborative filtering. These were demonstrated on a prototype web-site.

The team implemented a software system to help users find articles relevant to their interests from news sources such as Usenet and Reuters. The user provides feedback to the system, indicating whether the documents found are relevant or irrelevant. The system uses this information to automatically construct new queries that find documents that are relevant to the user.

The team developed a registration tool for radiographic images to be used in patient alignment process at the Loma Linda University Medical Center's proton therapy facility. The tool provides a quick, accurate, and robust registration of marked digitally-reconstructed radiograph (DRR) images taken in the planning stage of treatment with unmarked x-ray images taken during the patient alignment, allowing for a faster and more accurate patient alignment. The tool also gives the technician a clearer view of the patient alignment by reducing noise and removing artifacts found in the x-ray and DRR images.

The team developed a user interface which will be integrated with IBM's Grand Central Station (GCS) project, a data harvester coupled with a profile engine which crawls the internet and intranets, searching for data that matches queries stored in the user profiles. Data collected by GCS is pushed to the Higher Ground system, and from there into various clients the team developed, including a Lotus Notes client, a PalmPilot client, and an Email client. Relevant technologies include XML, CDF, Tuple spaces, and push-based delivery.

The team investigated ways to detect potentially harmful messages in a computer network, then implemented two detection systems. One system uses software tools to collect information about the network and correlate message sequences with the suspicious circumstances under which they arise. The other system uses a custom packet snooping and demultiplexing system to detect attacks.