Astronomy
There's a reason my electives kept circling back to physics. I'm a member of RASC, and I take amateur astrophotography with a Seestar S50 whenever the sky cooperates. The night sky is the original messy real-world dataset: partial observations, instrument noise, models that almost work. The N-body simulator started as a coursework assignment and became a long-running side project because the math is honest in a way most software isn't: when energy stops being conserved, you've shipped a bug.
Fishing
Three things fishing teaches that no codebase will: patience without a progress bar, comfort with non-determinism, and the discipline of doing nothing while observing carefully. I fish lakes in Manitoba most summers. The relevant skill is the same one that makes a good engineer: pay attention to the small evidence the system is giving you before you change anything.
Gardening
Gardening is the longest-feedback-loop interest I have, and that's the point. You can't iterate weekly. You set a system up, you tend it, and you find out in months whether your assumptions about light, soil, and water were correct. It's the closest analog to designing a long-lived piece of infrastructure: most of the work is invisible, and the failure modes are slow.
