Parameter inference in ordinary differential equations is an important problem in many applied sciences and in engineering, especially in a data-scarce setting. In this work, we introduce a novel generative modeling approach based on constrained Gaussian processes and use it to create a computationally and data efficient algorithm for state and parameter inference. In an extensive set of experiments, our approach outperforms its competitors both in terms of accuracy and computational cost for parameter inference. It also shows promising results for the much more challenging problem of model selection.