Tell me what you’re building.

A project, a pilot, or private on-prem AI, the form below lands directly in my inbox. No autoresponder, no inbox routing. I read every message and reply myself.

What I work on

Most of the work is private AI that runs on hardware the client controls: retrieval over internal documents, agents that touch real systems, and the plumbing that keeps both auditable. It tends to start with a dataset nobody trusts yet, or a manual process that has outgrown the person doing it.

Projects run from a short scoping exercise to a working pilot. If the honest answer is that a smaller tool or an off-the-shelf product solves it, that is the answer you get.

Finished examples, with architecture and stack, are on the portfolio page. The blog covers how these systems are built and where they fail in production.

What to write about

Good fits are a private AI pilot on your own servers, a private cloud or a GPU box you already own, retrieval over internal documents with cited answers, an agent that has to touch a real system, or an LLM feature that works in the demo but not yet in production. If fine-tuning is on the table, say what you have already tried with prompting and retrieval. Include the systems involved, roughly how much data there is, and whether it may leave your network. Questions about a portfolio project or a blog post are fine too.

What happens after you write?

I read every message myself and reply personally. Direct email isn’t published, which kills the spam loop. If it’s not something I can help with, I’ll say so; either way you get a reply.

It helps to mention what the system should do, where the data lives today, and whether everything has to stay on your own infrastructure.