We consider the problem of estimating parameters and unobserved trajectories in nonlinear ordinary differential equations (ODEs) from noisy and partially observed data. We focus on a class of state-s