SEREN: A Differential Privacy Application for Networking

Conference attendees are frequently overwhelmed by the scale of the event they participate in. Young participants, most especially find it difficult to identify the right co-attendees to reach out and connect with. We believe that by creating a simple and intuitive app that removes the anxiety and uncertainty from arrangement of such meetings we will dramatically increase the quantity and quality of connections. Following each match-making we collect user feedback and retrain our model to improve the quality of recommendations before the next round. The goal is for users to report that through the usage of this app they were able to make more meaningful connections during the conference than they would otherwise be able to. We used a differential privacy approach in order to protect users’ data in training the model as opposed to a regular machine learning approach. Poster / Presented and won MIT press Award at Deep Learning Indaba