Building Math tasks recommendation algorithm using CLIP model
12:15 - 12:45
CLIP is a new model created by the OpenAI team, and it allows us to generate images by describing them. In the recipient project, we had the task of creating a recommendation engine to optimize the number of successfully solved mathematical problems in a marketplace. My team has created a novel approach to the math-problem recommendation system that uses CLIP architecture and the training process. Our solution was able to find pairs between mathematical problems and their solutions to create vector representations of both tasks and math experts, and finally, offer them a sequence of unsolved tasks in an order that optimizes KPIs set by our company. This session will explain how the original CLIP model works and our solution to the recommendation problem.