We present a two-phase method for large-scale nonlinear constrained optimization that can readily take advantage of iterative subproblem solves. In the first phase, projected-gradient iterations approximately minimize the augmented Lagrangian to estimate the optimal active set. In the second phase, we solve an equality-constrained quadratic optimization problem to obtain fast convergence. We employ an augmented Lagrangian filter to determine the accuracy of the iterative augmented Lagrangian minimization and to promote global convergence. Our algorithm is designed for large-scale optimization, and we present preliminary numerical results.