Much of academic research and problem solving happens in clustered communities; which reside in bubbles of thoughts, ideas which rarely interact. CrowdAI is an attempt to connect numerous groups of academic researchers to a community of machine learning and data science enthusiasts. Academic researchers from various sub-domains of science act as problem owners, who bring a novel problem to the table; and CrowdAI works with them to strip away the domain specific information and make the problem both appealing and approachable to a community of machine learning enthusiasts. In the process, the group of academic researchers have access to state-of-art ideas from Artificial Intelligence; and the community of Machine Learning enthusiasts get exposed to a wide variety of problems that they would normally never get to work on. This talk discusses some of our recent success stories, and the lessons we learnt from them.