This is quite a big question. Do you mean how to build a structure from the data? You can use TALOS to predict the backbone torsion angles from the chemical shifts. Coupled with sufficient distance restraints from e.g. NOEs and residual dipolar couplings it becomes possible to then calculate a structure. CYANA is used for the final model building.
You might find this article helpful: https://www.sciencedirect.com/science/article/pii/S2001037017300053
At my site, people often use the SANDER MD engine from the Ambertools for that. The usually start multiple simulations from an unfolded conformation, pulling the NMR based couplings together. Here's a tutorial: