A model like ChatGPT is the starting point, not the finish line. My work is everything that turns that starting point into an AI that remembers, acts, stays safe, and solves a real problem. Here's what that looks like — in five plain-English capabilities.
Out of the box, AI forgets you the second a chat ends. I build memory systems so it can remember who you are, your preferences, past decisions, and things it figured out earlier — even weeks later.
Some of my memory designs are even inspired by biology: facts get stronger the more they're used and gently fade when they're not — just like human memory.
Chatting is nice; doing is better. I build AI "agents" that can break a big request into steps, use tools (search the web, edit files, send messages), and carry the work through to the end — with your sign-off where it matters.
The pattern I use: the AI classifies what you want, plans the steps, executes them, and checks its own work.
Many of my systems run entirely on your own computer using local AI models. That means your documents, messages, and data never have to be sent to any outside company — a huge deal for anyone handling sensitive information.
When cloud AI is the right call, I design it so you stay in control of costs and what's shared.
Powerful AI needs guardrails. I build in approval steps (nothing important happens without your click), strict limits on what the AI can touch, and complete audit logs so every action is recorded and reviewable.
You should never have to wonder what your AI did. With my systems, you can simply look.
This is the "architect" half of the job. Before building, I plan how memory, decision-making, safety, and the model connect — so the finished system is fast, reliable, and easy to change later. More on architecture →
Some of my most advanced systems can notice a gap in their own abilities, build a new tool to fill it, test that tool safely, and keep it if it works — getting more capable over time, on their own.
I'm an AI developer (I build the software) and an AI architect (I design how it's put together). Together, that means I take the powerful-but-raw AI models that exist today and turn them into complete, trustworthy systems that actually help people.
A great chef doesn't grow the wheat or raise the cattle — they take excellent ingredients and turn them into a meal worth serving. I do that with AI: I don't build the raw models, I turn them into something worth using.
Anyone can bolt a chat window onto an AI model. The interesting — and hard — work is everything below the surface. That's where I focus.
I build systems meant to run for hours or days, not flashy one-minute demos that break under real use.
I borrow ideas from living things — memory that fades, energy that must be managed, growth over time — to make AI behave more naturally.
Transparency and approval are built in, not bolted on. The human is always the one holding the wheel.
The architecture page shows how memory, decisions, and safety fit into one working system — with simple diagrams and analogies.