In this talk I present a new algorithm for the optimization of convex functions over a polyhedral set. The algorithm is based on the spectral projected-gradient method, but switches to quasi-Newton iterations whenever possible. A practical application of the framework is the Lasso problem, which also appears as a subproblem in the basis-pursuit denoise solver SPGL1. Other important applications that could benefit from the proposed algorithm include bound-constrained optimization and optimization over the simplex.