I build AI systems that need to work in the real world.

I studied advertising, but long before working professionally in technology I was already spending a lot of my time experimenting with computers, software, and hardware.

By the time I was 13, I was already installing homebrew on PSPs and iPods and modifying devices out of curiosity. Later on I started building PCs, experimenting with Hackintoshes, and getting increasingly interested in both hardware and software. I had no idea I would eventually work in tech; I simply liked understanding why something worked, why it stopped working, and how to fix it.

Then came crypto. A while before Ethereum really took off, I was already running a couple of mining rigs. Over time, that grew into almost 10 rigs with around 50 GPUs in total. I built the machines, replaced components, troubleshot failures, and managed the cluster remotely. It was probably my first experience maintaining infrastructure that actually needed to keep running.

AI followed a similar path. I started by experimenting with it and eventually found myself building systems around LLMs, agents, RAG, vision, document processing, model evaluation, and local inference.

That background advertising on one side, self-taught infrastructure and hardware on the other, this shapes how I build today. I care just as much about whether a person understands a screen as I do about whether a model fits into 24 GB of GPU memory.