Although quantum mechanical models have been applied to molecular dynamics for decades, the immense computing power now available makes this feasible on an unpredecented scale. By using an ONIOM-based approach that combines density-functional theory and Stefan Grimme's xtb semiempirical methods, it's now possible to simulate large systems (over 1000 atoms in the "low" level) for tens of picoseconds. This approach can be combined with existing biased-sampling techniques like WHAM, enabling the construction of solvated free energy surfaces for complex transformations, or used to study the lifetime of reactive intermediates.
I’ve used AIMD to address tricky computational problems in a few projects, listed below. With Eugene Kwan, I also develop and maintain presto, an open-source Python package for running QM/QM’ calculations. For more details about presto, see the website.