2D nav mesh generation and agents with dynamic avoidance would be invaluable for a lot of people, especially if integrated with existing physics systems such as Box2D. Path finding solutions of the nav mesh could be abstracted into modules to allow reuse for other purposes (A* in particular can be used for a variety of problems).
But as this is also an academic project start by identifying an acceptable minimum feature set, and focus on that first. It is far more valuable to the community here to have well tested, stable utilities than a wide feature set riddled with bugs.
Start by researching other engines (UE4, unity, game maker) to identify paradigms and API people are used to using, as well as good archetectual patterns that will likely save you pain and time in the long run, as well as give you lots to write in your dissertation