Pick a project small enough to finish
The single biggest reason first robot projects die isn't a bad motor choice or a fried H-bridge, it's scope. Learners who've just finished courses on actuators, embedded compute, and power systems often want to prove they've absorbed all of it at once: a robot with six degrees of freedom, computer vision, autonomous mapping, maybe a robotic arm bolted on top. That robot is a fine goal for six months from now. It is a terrible first project, because when it doesn't work (and it won't, on the first try), there are so many interacting subsystems that you can't tell which one is broken. A small failure surface is what lets you actually finish something.
A differential-drive rover, two motors, two wheels, a caster or skid for the third contact point, is the right scope for a first robot. It's mechanically forgiving (no steering linkage to align), electrically simple (two motor channels instead of four or six), and the software is a single, easily reasoned-about control loop. Add one sensing task on top: either line-following with a couple of IR reflectance sensors, or basic obstacle avoidance with a single ultrasonic or IR distance sensor. Both give you a robot that visibly does something intelligent, and both are achievable with the parts and concepts you already have from courses 1 through 3.