This is a smaller project that I took on, but perhaps worthy of mention for its peculiarity. Every year, the University polls its faculty and staff to solicit volunteers for committees, asking for ranked preferences along with a number of demographic or position-related questions. I was asked to help make sense of this process, since the status quo was merely to gather all the results into a giant spreadsheet and begin moving volunteers around until it all seemed to fit.

After retooling the base survey, I loaded the results into MongoDB and then built a simple app that helps to query, navigate, and assign the volunteers to all of the committees, tracking the required compositions (e.g. “this committee requires 1 faculty rep from each college, and at least 5 professional staff”). Due to making the various aspects of the volunteer easy to track and make sense of (from service years, to position, etc), the application was recieved with immense gratitude by the committee that oversees committee assignments, and expanded a bit over time.

An interesting side project within this application was a journey into optimization. I recognized at one point how easy it is to inadvertently make committee assignments that block the possibility for as many people as possible to attain one of their highest designated preferences. I used Google’s OR-Tools to write an optimizer that scans the current assignments and the list of remaining volunteers and the committee requirements, and then computes what percentage of persons will be able to reach their 1st preference, 2nd preference, etc. I assigned weight to those cases as well, in order to produce an overall “cost” score to indicate how well the assignments were made. This mechanism allowed one to see the cost added when selecting someone for a committee, which might lead to hesitation if it is very high. I wish to expand this use to build in cost functions around other measurements (for instance, penalize having a committee populated by faculty primarily from one set of disciplines rather than dispersed, etc).