Who's Behind This
Jim Fischer
I spent two decades building and securing enterprise infrastructure before AI was a buzzword. Now I help small and mid-sized businesses put AI to work on hardware they own, with the cloud as a deliberate exception instead of the default.
The Background
My career started in enterprise network and security infrastructure — the kind of work where a misconfigured firewall meant a front-page incident. I worked with DEC and Compaq Professional Services deploying and hardening infrastructure for organizations that couldn't afford mistakes.
In 1999 I completed a residency at Microsoft's data centers, deploying AltaVista Firewalls in high-availability configurations. That era taught me something that still holds: the most dangerous assumption in security is that someone else is handling it.
After that I spent years in ISP network operations, then transitioned into precision CNC manufacturing — 5-axis programming, GD&T, fixture design. That background matters here: I think in systems, tolerances, and failure modes. I don't deploy AI for the sake of deploying AI.
When large language models matured enough to run locally on affordable hardware, I built my own private AI infrastructure from the ground up. RAG pipelines, vector databases, local model serving, network-segmented deployment. I know exactly what it takes to make this work in a real environment.
Why Local-First
I'm in this because AI is useful. It turns an afternoon of digging through files into a question and an answer, and a blank page into a draft worth editing. Local-first is how a business with sensitive data gets that without handing the data to someone else.
Consumer AI tools may train on what you type. As of October 2026, the business tiers of ChatGPT and Microsoft 365 Copilot contractually don't. That's a real protection, and it isn't the same as your data staying in your building. Your prompts and files still sit on their servers, under their retention rules and their uptime terms.
Local-first AI keeps that data on your hardware. The model runs on your network, and nothing leaves unless you've decided a specific task should. You control the updates, the access, and the audit trail.
Local open-weight models don't match the best cloud models on the hardest reasoning, and I say so on every first call. What I build is different: your firm's knowledge, meaning your documents, your workflows, and your constraints, runs on hardware you own. For the occasional task that needs frontier reasoning, we agree on what may leave and under what rules, and I build that path during implementation.
This isn't paranoia — it's the same logic that drives on-premise ERP systems, air-gapped networks, and private data centers. Some data is too sensitive to hand to a third party, no matter how good their marketing is.
How I Work
Scope First
I won't recommend hardware or software until I understand your actual workflows. The audit exists so we both know what we're building before anyone spends money.
Fixed Fee
Every engagement is scoped and priced in writing before work begins. The fee covers that scope. Anything outside it gets a written change order before I do it, so there are no surprise invoices.
You Own It
You own the audit deliverables once they're paid for, and your data stays yours. Anything I install runs on open-source software and open-weight models on hardware you own, with no license fee to keep it running. I keep my methodology, templates, and internal tooling.
Direct Communication
I tell you what I find, including the things that are inconvenient. If something won't work, I say so before you've paid for it. If your data can safely live in Microsoft 365 Copilot or ChatGPT's business tier, those are cheaper and smarter than anything I'll install, and I'll tell you so on the first call.