about
I started in hardware — the kind of engineering where your mistakes have physical consequences and debugging means grabbing an oscilloscope. There's something humbling about working at that layer. Electrons don't lie, and abstractions don't save you.
Over time, the stack shifted upward. I moved into machine learning and data science, drawn by the same core instinct: take a messy, poorly understood system and extract something useful from it. The tools changed, but the disposition didn't. I still think about problems from the substrate up — which is a strange and occasionally useful way to approach an LLM pipeline or a data model.
These days I spend most of my time at the intersection of data, AI, and building things that actually work. I'm interested in agentic systems, the messiness of real-world ML, and the craft of turning ambiguous problems into something you can ship.
This blog is where I think out loud — rough ideas, half-finished experiments, things I'm figuring out as I go. Hence the name.