At the end of the project, we will have a TA-supervised evaluation in which each submitted die is rolled a large number of times to estimate the face probabilities. (How many rolls are necessary to have good estimates? Do we need to be careful about how a die is rolled?) Make sure your dice are robust enough to survive many hundreds of rolls.
We'll have several categories in which entries can be submitted, most of which we will decide collectively as the class discussions progress. One category that will definitely be included is: Build a 6-sided die aiming for the minimum empirically-measured discrepancy.