Before I explain what I do, it helps to clear up what artificial intelligence actually is. No technical background needed. If you can follow a recipe, you can follow this.
AI is software that learns patterns from huge amounts of examples, and then uses those patterns to make good guesses about new situations. That's it. It isn't alive, it isn't conscious, and it doesn't "know" things the way you do. It's an extraordinarily good pattern-matcher.
Think of how a child learns what a "dog" is. Nobody gives them a rulebook. They just see hundreds of dogs — big, small, fluffy, spotted — and eventually they can spot a brand-new dog they've never seen before. AI learns the same way, except it looks at millions of examples instead of hundreds.
You'll hear names like ChatGPT, Claude, or Llama. These are AI models — the raw "brains" trained on enormous amounts of text. A model on its own is like a brilliant expert locked in a soundproof room with no memory, no hands, and no way to check the time. It can answer a question you shout through the door, but the moment the conversation ends, it forgets you completely.
That gap — between a raw model and something genuinely useful — is where AI development lives.
AI development is the work of building everything around the model so it becomes a real, dependable tool. It answers questions like:
Answering those questions well is a craft. Doing it in a way that's reliable, secure, and pleasant to use is the difference between a toy and a tool.
An AI model is an engine. Powerful, but useless sitting on a garage floor. AI development is building the whole car around it — the steering wheel, the brakes, the fuel system, the dashboard. My work is designing and building that car so you can actually drive somewhere.
If AI development is building the car, AI architecture is being the person who designs how all the parts fit together before anyone starts building. An architect of a house decides where the rooms, plumbing, and support beams go. An AI architect decides how memory, decision-making, safety, and the model itself connect — so the finished system is fast, reliable, and doesn't collapse under pressure.
It's the difference between a pile of great parts and a machine that actually works together. There's a whole page on this →
| Term | What it actually means |
|---|---|
| Model / LLM | The trained "brain." LLM means "Large Language Model" — a model that works with words. |
| Prompt | The instruction or question you give the AI. Like the words you'd say to a very literal assistant. |
| Agent | An AI that can take actions on its own — not just chat, but do things step by step. |
| Memory | A way for the AI to store and recall information between conversations. |
| Local / on-device | The AI runs on your own computer instead of a company's servers. More private. |
| Cloud | The AI runs on a big company's remote computers. Convenient, but your data leaves your device. |
| Training | The one-time, expensive process of teaching a model patterns from examples. I don't retrain models — I build with them. |
| Fine-tuning | Gently adjusting an existing model to be better at a specific job. |
People often think AI development means "creating a ChatGPT." Almost never. The models already exist and are made by a handful of giant companies. The real, valuable work — and what I do — is turning those models into systems that solve actual problems: that remember, act, stay private, and can be trusted.
That's the whole game. And it's what the rest of this site is about.
Ready to see how I put these ideas to work? Here's what I actually do, day to day.