« home

TeamMail - Enhanced Email for Collaboration

For my capstone project in my HCI Masters program, my 5 member team, under the sponsorship of Sun Microsystems, designed and developed an email system for better supporting collaboration. They gave us an open ended challenge to find collaboration breakdowns and make a system to fix them, and we chose to focus on the integration of tasks with email.

We did an amazing job if I do say so myself, and I encourage you to look at the TeamMail site.

LJNet - LiveJournal Social Network Browser

A common problem with social network software is that for social purposes, they have not given any functionality to end-users beyond the random wandering through linked profiles. I believe social network applications will be far more useful with effective visualizations and interaction techniques, and to that end I developed a system to explore the social network of a LiveJournal member. This social network browser (that I will soon release publicly) enables users to effectively see overall patterns in the connectivity and common interests between their friends and friends of friends - but showing is much better than explaining, so please visit my website for LJNet.

Socializer - Mobile P2P Social Platform

When I worked for IBM during summer 2003, they asked my 4 person team to design and build a P2P 'killer app' to push market adoption of the Java OSGi standard (dynamic services for networked devices). We created the Socializer platform by turning the existing server/client model of OSGi operation into a P2P, distributed system for PDA's and laptops. By detecting other users of Socializer in the vicinity, our platform supports pervasive mobile social interaction with both people and services via plugin applications (like chat, file transfer, etc) that can be shared in a P2P manner - you can see the various plugins being used by those around you and with a single click download them and then be able to interact in a new manner.

The interface shows users (1) who and what services are around them that they can interact with, (2) what plugin applications are currently installed, and (3) what interactions they currently have open. For example, Josh might open up his client and see Heather around (they may or may not know each other). By tapping on her name, he sees what interactions are supported by the applications they have in common (the default installation comes with the applications shown here). Much more information can be seen on the IBM alphaWorks download site.

AMusic - User-Empowering Music Recommendations

Following my research into Collaborative Filtering systems (like Amazon, audioscrobbler, gnod, movielens, etc), I saw that some clear problems with all such systems were their limited functionality and usability. This led me to design and build a music recommendation web app that did not require a lengthy user profiling stage to 'learn about their interests' and secondly had the functionality to refine system inputs and navigate the recommendation space. Basically I made a music recommender that (instead of just giving recommendations as ultimatums to users) used collaborative filtering to support a music recommendation search engine - so that searching becomes as quick and simple to use as a google search, and returns vastly more navigable and customizable search results than existing recommendation systems.

I haven't yet resurrected the system, but I can show a couple screenshots of the opening screen and the navigable search results page. The database of music connections was populated by using the google API's to find music playlists (generated by mp3 playing applications) that had been posted online, and analyzing the artist and track information across the 1,000 or so of them that I could get (over 400,000 individual songs).