Neural quantum states (NQS) provide a powerful framework for representing and simulating quantum many-body systems, offering an alternative to traditional methods like tensor networks. In this talk, I will give an introduction to NQS, explaining how neural networks can encode complex wavefunctions efficiently and highlighting their applications in ground-state search and state reconstruction. I will then delve into using NQS for time evolution, focusing on a time-dependent neural quantum state (t-NQS) approach. Within this method, a neural network represents the quantum state across all time steps, with time as an explicit input parameter. Unlike conventional step-by-step integration methods, this approach frames time evolution as a global optimization problem, leveraging automatic differentiation for efficient training. I will discuss the advantages of this formulation, as well as provide an outlook on ongoing projects that build on this method, exploring its potential applications and future directions.