In line with this question, I'm looking for suitable functionals for transition metal compounds with organic ligands to predict NMR spectra. Also, I'd like some guidelines about how to choose functionals for this type of calculations.
I haven't had much experience with calculating chemical shifts, but I do have experience calculating J-coupling between P atoms on metal complexes.
From this, I know that for chemical shifts and J-coupling calcs may require the use of specialized basis sets since most energy optimized basis sets (i.e. pople, dunning, karlsruhe) do not have an appropriate description of the core electrons for accuracy in nuclear calculations. I don't have experience with basis sets that are developed for chemical shifts, but this paper may be of use (although it only looks at 31P shifts): https://dx.doi.org/10.1021/acs.organomet.0c00127
I know that for 13C and 1H chemical shifts, the hybrid GGA functionals WPO4 and WCO4 were developed by Cramer for small organic compounds (ref to the original paper: https://pubs.acs.org/doi/pdf/10.1021/ct6001016 ) and are available by default on many quantum chemical computing programs - the paper cited also discusses the performance and shortcomings of other functionals, so that may help guide you if WPO4/WCO4 are not appropriate for your system.
Also the GGA functionals KT1, KT2, and KT3 were developed for the nuclear shielding constants of non-H atoms and could be useful to you too (ref to the original KT paper: https://doi.org/10.1063/1.1590634)
I do not have much experience with DFT, but have been digging into DFT in strong fields the last couple of months. The investigation of magnetic properties such as NMR shielding constant and the magnetisability in the context of DFT rely on an explicit dependence on the magnetic field. Magnetic-field DFT, known as BDFT (J. Chem. Theory Comput. 2017, 13, 9, 4089-4100) and current DFT (CDFT, J. Chem. Theory Comput. 2015, 11, 4169−4181) are alternatives to the the standard DFT, where it was shown that cTPSS functional gives modest improvements over GGA level functionals such as PBE.
Hence, you might try to use cTPSS functional and its hybrid forms, i.e. Hybrid-cTPSS (10% HF exchange is added) and Range-separated cTPSS.