I build products with AI inside them, not demos.
AI features wired into real products — assistants, automation pipelines and browser tools built on models rather than around them.
AI features built into a product rather than bolted onto it
Model integration — OpenAI and comparable providers
Automation pipelines that run without someone watching them
Chatbots and assistants with the context to be useful
The interface around the model, which is usually what decides whether it works
Define
What the AI is actually for. Most AI features fail because they answer a question nobody asked.
Prototype
The smallest version that proves the idea, built fast enough to throw away if it does not.
Build
Model, pipeline and interface together — the wrapper is as much of the product as the model.
Ship
Deployed with sensible limits and fallbacks, because models fail differently from software.
- OpenAI GPT-4o
- n8n
- Convex
- ElevenLabs
- JavaScript
- Chrome Extension
Do you train your own models?
No. I build products on top of existing models — OpenAI and comparable providers — and the work is in the pipeline, the context and the interface around them. For almost every product, that is where the value is.
What kind of AI work have you done?
PresenceIQ, an AI avatar platform with a pre-conversation intelligence pipeline built in 24 hours using Convex, n8n, GPT-4o, ElevenLabs and BeyondPresence. And AIX Lens, a Chrome extension that flags potential misinformation on a page in real time.
Can you add AI to a product we already have?
Yes, and that is usually the more sensible starting point — one well-chosen feature inside an existing product beats a new AI product with no users.
How do you handle cost and reliability?
Model calls are metered and cost money, so they get limits, caching where it helps, and a defined fallback for when the model is slow or unavailable.