Things are going to differ, slightly, for quant finance and data science. For quant projects, I’d advice to not work on something that is hip and full of fluff (things like deep learning, etc.) but work on simple stuff rooted in the economic implication of what you are analysing.
On things you can explore one could be NLP for analysing textual data like analyst reports, financial news articles, financial statements & disclosures, and maybe even tweets (be careful this with one since it can have a lot of noise). Another interesting thing would be to look into ESG data.