I'd say it's definitely a very nice idea for an academic project. My book recommendation is just Mastering the Lightning Network (https://github.com/lnbook/lnbook).

It's unfortunately not as simple as you imagine. Trick is your lightning node implementation already does a lot of trickery to chose the optimal path gauging between reliability and price. I mean there is always some room for improvement but the low hanging fruit is gone I am afraid.

What you could do is say ok my payments always go through that route. If I connected to that node directly I could leave out all those nodes and save that much on routing fees. But opening a new channel is an investment. You don't know how the network will evolve in the future. Perhaps the nodes you planned to skip will drastically lower their fees and you actually wasted money on transaction costs. So you also can't make such a solution commercially viable (since you cannot tell a user I saved you 100k sats and now I will take my cut of 10k sats).

Perhaps you could charge per advice. Sort of like investment advice. But OTOH we all know that bitcoin is the best investment and similary the best advice is simply to open a channel if you have a real life connection with the entity. For example if the other node is a bar where I drink it makes sense to open a channel because I am going to pay mulltiple times. But if somebody just says hey this node is the best based on this criteria this is more like stock picking. Ok somebody did some analysis so it's probably better than random choice but it's still something that might never pay off.