We designed a domain-specific programming language for modeling the workflows of media publishing, and implemented a compiler for realizing those workflows. Our programming language allows media distributors to concisely specify their business logic for bringing media from internal production to Internet release. Our compiler translates that specification into tools that use the Platform's products to implement that logic.

Providing large, networked services involves the operation of many smaller components. A network management system can help quickly diagnose problems when a component fails. In a large network, it is useful to distribute work across several smaller networks, each with their own network management system. The Configuration Management Database Federation (CMDBF) defines a web service to allow communication between different data repositories. We set out to extend Zenoss, an open source network management system, to provide the CMDBF web service.

ODE Toolkit is a program for solving and teaching Ordinary differential Equations. Over the past five years, CODEE has employed students at Harvey Mudd as developers. Unfortunately, weak architectural foundations and insufficient documentation have led to an unwieldy code base and significant development overhear. Our primary goal is to apply standard software engineering principles to reduce the burden of maintenance and upgrading. Additional, the code will be prepared for open-source release to encourage the use and development of the ODE Toolkit.

Magnify360 provides the most sophisticated marketing optimization platform for the modern Web. Featuring Predictive DNA™ technology, magnify360 dramatically outperforms traditional conversion optimization tools by personalizing each visitor's browsing experience in real-time. The self-learning system immediately adapts Web site content, layout and page flow based on an individual's behavior on-site, ultimately driving them toward a specified call to action. The goal of this project was to investigate new algorithms as possible solutions for the next generation of the magnnify360 self-learning system.

In the era of cloud computing, many of us take our data for granted that once date has entered the cloud we consider it safe from loss. Companies providing cloud-based services work hard to try and maintain these guarantees, but many of their methods are based on things that "probably" work. We present Zab, a replicated transaction log library that provides safe data logging and guarantees consistency.

Genius.com provides e-mail tracking for sales and marketing professionals so they can track their customers' interests and keep in contact with them. Unfortunately, these tracking techniques cannot distinguish between the original recipient and those to whom the e-mail was forwarded, which leads to confusion when the marketer contacts the customer based on this mixed information. We developed a decision tree algorithm to distinguish between different e-mail recipients in real-time and a user interface to display our results.

The MySpace social networking website allows members to create personal profiles and share them with other users. One significant feature of MySpace allows members to share their tastes in music and to listen to popular music. The goal of this project is to use this data to develop and assess a prototype music recommendation system for MySpace. We have created an application that reads a user's music preferences from their profile and recommends other artists the user may appreciate.

The team has prototyped an automated system for increasing collaboration in corporate environments. The system analyzes corporate email to discover opportunities for collaboration. Employees then receive recommendations to communicate with other, similarly interested people. The project leverages machine learning, clustering, and text processing techniques.