Technical résumés love to list tools like a shopping receipt. That tells you nothing if you're not a programmer. So here's my toolkit grouped by the job it does, with each item explained by what it actually accomplishes.
Tools change every year. These underlying abilities are what really matter — and what I bring to every project.
Seeing how all the parts of an AI system should fit together — and planning it before building, so it works reliably.
Designing how AI remembers, forgets, and recalls — including memory modeled on how the human brain works.
Building AI that plans and completes multi-step tasks on its own, safely and within limits.
Getting several AIs to collaborate — debating, dividing work, and checking each other.
Creating living virtual worlds and characters with goals, moods, and behavior over time.
Running AI locally, guarding data, and adding approval and audit controls people can trust.
Programming languages are just the "vocabularies" I use to write software. I work primarily in two:
| Tool | What it's for, in plain English |
|---|---|
| Python | The go-to language of AI. I use it for the "thinking" parts — memory, reasoning, simulations, and connecting to AI models. |
| TypeScript / JavaScript | The language of the web. I use it to build the screens, dashboards, and 3D worlds people actually see and click. |
I design systems to work with many AI models, and to swap between them freely — so you're never locked to one provider.
| Tool | What it's for, in plain English |
|---|---|
| Ollama (local models) | Runs AI models directly on your own computer — private, offline, no subscription. A cornerstone of my privacy-first work. |
| Claude, GPT, Gemini, Grok, Mistral, DeepSeek | The leading cloud AI models. I connect to whichever is best for the task — and let you switch anytime. |
| Provider-agnostic design | My systems treat the AI model as a swappable part, so upgrading to a newer model doesn't mean rebuilding. |
| Tool | What it's for, in plain English |
|---|---|
| Databases (SQLite) | The "filing cabinet" where an AI stores what it remembers so it survives being switched off. |
| Knowledge graphs | A way of storing facts as connected, readable notes — so the AI's knowledge can be inspected and corrected, not hidden. |
| Vector / semantic memory | Lets AI recall things by meaning, not exact words — so it finds "that thing about my car" even if you phrase it differently. |
| Tool | What it's for, in plain English |
|---|---|
| React | The tool for building clean, responsive screens and dashboards. |
| Three.js | Creates interactive 3D worlds that run right in a web browser — no download needed. |
| Text-to-speech (Edge TTS, ElevenLabs, and more) | Gives AI a natural spoken voice, so systems can talk instead of just type. |
| Web Speech / voice input | Lets you talk to the AI out loud, hands-free. |
| Tool | What it's for, in plain English |
|---|---|
| FastAPI | The "switchboard" that lets the different parts of a system talk to each other quickly and reliably. |
| Discord & Telegram integration | Lets your AI live where you already chat, so you can reach it from your phone. |
| Approval gating & audit logs | Safety features I build in: nothing important happens without a click, and everything is recorded. |
| Encrypted credential storage | Keeps passwords and keys locked away using the same protection your operating system uses. |
For the technically curious — everything above, condensed.
See these tools and abilities working together in real, finished systems.