I keep ending up at the intersections. A routing algorithm for pharmacy trucks turns into a reinforcement learning problem. A proxy cluster that has to survive hostile networks fails the same way trading infrastructure does. I’ve worked on robotics, ML, embedded systems, finance, and honestly the most useful thing about any of them was how they made me better at the others.
The question I come back to isn’t “how do I optimize this?” It’s “what is actually hard about this problem?” Most complexity is inherited. Someone made a decision three layers down and everyone after them worked around it. I’d rather find that decision and revisit it.
That’s where I want to keep working. The places where fields overlap and the interesting problems haven’t been properly named yet.