Journal Entry - Spring 2026

This semester, I got to take Circuits 2 with Dr. Sheplak, a course I REALLY REALLY enjoyed. The techniques and intuition we learned are invaluable, and will undoubtedly apply for my entire career. Bode plots were an unintuitive mess before this course, as were Thevenin/Norton Equivalents. Not the case anymore! Also got to take EEE3308C, where I got to learn about MOS circuits, Op-Amp internals, and the design tradeoffs that go into all such circuits, although our coverage of these topics was surface-level. I also got to be a PI for both EEL3701C and EEL4744C under Dr. Schwartz, where I got to teach many students the underlying intuition of digital design concepts, how to break down problems, and the value of mixed logic circuits in reducing complexity. I was also a member of the Machine Intelligence Laboratory (MIL) this Semester, where I designed the leak detector for the Sub. I am now working with Mehran in the FOCUS Lab under Dr. Koppal. This semester we built EdgeTrain, a runtime framework that lets mobile robots learn continuously on edge GPUs without overheating. The core discovery is that different continual learning algorithms heat the chip in measurably different ways, so algorithm choice itself becomes a thermal control knob alongside GPU frequency and batch size. We validated it on a Jetson Orin NX under an infrared heat lamp, showed that thermal throttling doesn't corrupt learned weights.