If you're making Western classical music in classicist, romanticist or modernist style, the model of music you have will carry a lot of information about harmony and the application of harmonic techniques throughout the piece. Given a core musical idea you can then apply peripheral techniques (such as orchestration) to build a full piece. E.g. when people study counterpoint, the model of music originates from vertical harmony of notes and when they can be used with respect to each other. The assumption is that orchestration is something that'll be separately developed "skinning" the composition. E.g. a common technique in this tradition is composing a piece for piano four hands and then orchestrating it (e.g. Holst's "The Planets" symphony was composed this way).

However, this stops being a useful model once you step into other musical traditions. In some cultures harmony would be like how Western music treats orchestration, peripheral to composition (like how extreme speed is irrelevant to Newtonian mechanics because it was never designed for near lightspeed motion). So you'd first design timbres, and have an idea about how timbres interact, timbres change, transform to each. You may have a theory of counterpoint of timbres. Once you have this, you can apply any standard "harmony skin" on the composition and you have a piece. This is not even restricted to non-Western music. If you look at the postmodernism in Western music you'll find instances of it. Easy example: a lot of people say that Philip Glass "makes the same music" again and again, what is being missed is the point he's trying to convey is that even if you pick the exact same 4 chords you can still create variation in music via other means. It just won't be different from the traditional harmony-centric Western musical model.

By the way, I studied CS and my full-time job is a Software Engineer. So I doubt our disagreement comes from my background in computer science.