I always feel like gathering data and formatting correct data sets is the most important part of data science. Back during classes where you're taught machine learning require perfect data sets for you to learn the algorithms. And then when you're analyzing data, you'll use a library that runs these algorithms for you. Which means you'll spend the longest time making sure that you're data sets are correct and easily useable.

I love working on these types of projects. Things like scraping sports stats (which may or may not be considered legal, but not like I'm trying to make money for them), data from posted articles, music data, data about locations of public restaurants, etc.

Finding some place that offers correct public data is fantastic to see.