This project develops a recommender system by solving the matrix completion problem and provides a comparison of different Optimization algorithms, such as the classical Frank Wolfe -and the pairwise variant-, and the Projected Gradient Descent. All three algorithms were trained and tested on three datasets with different levels of sparsity to obtain a richer outcome of the algorithms' performance.