I found my way into applied machine learning through hands-on engineering. I started as an electrician, moved into mechatronics and high-voltage battery engineering, and later into embedded systems, simulation, and data-driven modeling.
Batteries are my deepest domain, but my work extends beyond them. I am interested in time-series modeling for physical systems, especially when data is messy, compute matters, and models have to hold up outside a clean benchmark.
Outside work, I have spent long stretches traveling across Asia and Australia.
Why physical systems
I got into batteries somewhat by accident after undergrad, when I joined an 800 V/750 kW automotive prototype project. I stayed because batteries are hard in the right way: messy physics, real hardware, tight safety margins, and the scale at which small improvements matter.
The same challenge appears in many physical systems: observations are incomplete, operating conditions shift, and the model is only one part of a larger engineering loop.
I think of a neural network as a generic mesh and backpropagation as a generic solver. ML earns its place when it shortens expensive engineering loops or captures what explicit models miss. It becomes risky when the benchmark is cleaner than the real task, so I care about long-horizon behavior and realistic constraints.
How I work
I learn best by building and going down the stack until assumptions become visible. That habit carried me from electrical work into embedded systems and machine learning, and it still shapes how I approach unfamiliar problems.
I am comfortable working independently and owning a technical direction, but I do not treat engineering as a solo activity. The hardest problems cross disciplines, and progress depends on making trade-offs legible, listening to the people closest to each part, and turning uncertainty into shared experiments.
I prefer simple, inspectable tools and systems I can understand end to end. I care less about novelty for its own sake than about whether an idea survives contact with data, hardware, and the people who have to use it.