The projects aren't practice.

I've been into software and how it's built for years, long before I could make anything real with it. Computer science gave me the tools to finally build the systems I'd always wanted to, instead of just imagining them or selling someone else's.

My background is in B2B sales and business, and it didn't push me into engineering — the wanting to build was already there. What it did was tell me what to build. I spent years inside the workflows business software is supposed to serve: selling CRMs, selling construction tools to contractors, sitting with the people who used them all day. So when I had the skills, I didn't reach for generic practice projects. I built for problems I already understood.

That's the pattern for everything here. I don't build projects just to learn a technology; I build software for a problem I understand, and I learn whatever that problem needs along the way. My CRM and my estimator exist because I wanted the work to add up to something useful, not because I needed projects to show.

It's the same with the AI work. I've been building with LLMs since ChatGPT launched, and the lesson landed fast: the model isn't the hard part. Anyone can call an API and get a plausible answer back. The engineering is everything around it — gathering context automatically, standardizing the inputs, validating the outputs, chaining several model calls together, resolving dependencies, and keeping a person in control — so a probabilistic model behaves like dependable software. In my CRM the LLM is one small, carefully bounded step in a much larger workflow, and the workflow is the part I care about.