You still have a systematic increase. Statistical testing should be performed (at least simple block averaging) to see if it is meaningful. However, do not rely solely on RMSD to determine if you should stop your simulation. It is a degenerate, imprecise metric. You should look at ALL quantities of interest and determine if they have converged.

Justin, ideally what are the parameters/quantities shall I check in to see if my system has finally converged or not? If I look into physical parameters like temperature or pressure, they have converged properly, but I have usually used RMSD as the main parameter to see if my system has properly converged or not.

Anything relevant to the behavior and dynamics of your system. I'm not talking about temperature and pressure (which equilibrate fast, within the first ns or sooner). I'm talking about whatever physical properties are functionally relevant to your system. That's for you to decide based on whatever is of interest to you in doing the simulation. If those quantities aren't converged, you don't have reliable data.

Since RMSD is relative to another state, it is not very meaningful if the reference state isn't e.g. a crystal stucture rather than just the start point of the MD.

A better metric can be the radius of gyration. Which can be also plotted against time to conclude structural convergence. In VMD you could use the attached tcl script. Simply source it into the Tk Console, its interactive so will ask for everything. The resulting output file can be plotted with e.g. xmgrace.