There were many, one recurring one on the problem solving side: - you have databases and users who access them and everyone has access to everything, need to come up with an access model that will require some access permission while not introducing too much friction - on the stats and modeling side: Rotten tomatoes data, goal is to explore features that predict whether movie will be certified fresh or rotten by critics, data quite raw to turn into a model and iterate in 60 or 90 minutes, whatever it was