Private AI / Pittsburgh, PA

Your AI.
Your building.
Your data.

I build private AI for businesses that can't afford to have client data leave the office. Then I connect it to the tools your people already use, put the cloud AI you're paying for under control, and build agents that take real work off your team's plate.

Book a 30 minute demo on your files See the services
airman-ai / onsite appliance / sample LOCAL
> What does our records policy say about how long we keep closed client files?
ANSWER

Closed client files are retained for [X] years from the close date, then reviewed by the records manager before destruction. Files under litigation hold are excluded.

Records_Policy_v4.pdf p.3 Litigation_Hold_SOP.docx
model
on premises
sources
cited
cloud calls
none
The problem / Shadow AI

Your staff are already using AI. Probably three or four different tools, on personal accounts, with your client files pasted in. Nobody approved it. Nobody can audit it. And the first time it goes wrong, it's your name on the letter to the client.

Services

Four services. One path from first install to real automation.

Sensitive work stays on hardware you own. Everyday work can run on an enterprise provider. I set up and govern both.

01

Private AI, built and installed

A preconfigured AI system that lives in your office and answers from your files, with the source cited. No prompts go to the cloud. You own the hardware.

Sizing and build Onsite install Hardening Support
02

Connected to your tools and data

SharePoint, your file server, your CRM, your accounting system. Read only by default, every query logged, limited to what each person can already see.

MCP connectors M365 CRM SQL
03

Enterprise AI, managed

Already paying for Copilot, ChatGPT or Claude? I set it up right, lock down the data settings, shut off the unapproved tools, and report monthly on usage and cost.

SSO and seats Policy Spend reports Training
04

Agents that do the work

Contract intake. Document chasing. Request triage. One workflow at a time, a person approves anything that matters, and we measure the hours saved.

Workflow mapping Guardrails Approvals
The hardware / GPU inference

Real GPUs. Real inference. In your building.

Every system is sized to your team, built on workstation GPUs, burned in before it ships, and installed by me onsite. The models run on hardware you own.

A compact private AI appliance on an office desk
01 / THE APPLIANCE Sits in your office. Quiet enough for it.
A technician installing a GPU server into a rack
02 / GPU BUILD Sized, built and installed onsite.
A secured server corridor behind a vault door
03 / HARDENED No cloud prompts. No hidden data path.
computeWorkstation GPUs, sized per team
modelsCurrent open weight LLMs
retrievalLocal RAG, sources cited
refreshModels reviewed quarterly
How it works

It knows your documents. Nothing is uploaded anywhere.

An analyst working at monitors in front of server racks
inference: on premises index: your network cloud calls: none
  1. STEP 1 Connect Your approved sources: file shares, SharePoint, SOPs, contracts, tickets.
  2. STEP 2 Index A local retrieval layer, built and kept on your network.
  3. STEP 3 Ask Private model inference on hardware in your building.
  4. STEP 4 Answer Cited, auditable output your team can check in one click.
Built for regulated business

Who I work with

Based in Pittsburgh. Onsite across western Pennsylvania.

Machinist in a defense manufacturing shop CMMC L2 Defense suppliers AI that supports your CMMC program instead of breaking it. Attorney reviewing documents in a law office PRIVILEGE Law firms Client matters stay inside the firm. Records room with file boxes and a scanning workstation TAX SEASON Accounting firms Financial records never touch a public chatbot. Physician working at a desk in a clinic HIPAA Medical practices Patient data stays on hardware you control. Council chambers in a municipal building RIGHT TO KNOW Municipalities Public records work, without public cloud risk.
The three year math
42% less

Over three years for a 25 person team. And at the end, you own the hardware.

Cloud AI subscription$54,000
Airman AI private appliance$31,100

Cloud: 25 seats, about $18,000 per year total.
Private: $9,500 hardware and install, then $600 per month support.

Start small
90
day pilot

Prove private AI on one workflow that matters to you. If the numbers aren't there, you'll know in 90 days, not three years.

01
InstallI size, build and install the system onsite, then harden it and hand you the runbook.
02
ConnectWe pick one high value workflow and connect the sources it needs, read only by default.
03
MeasureWe agree on the success metric on day one and report against it at day 30, 60 and 90.
Why me

When you call, you get me.

I'm Kevin Vettorel. I spent twenty years in the United States Air Force running mission critical IT, and I've been building and deploying AI since long before it was on every billboard.

[Headshot]
30 yrsIT and security operations
20 yrsUnited States Air Force
9 yrsBuilding and deploying AI
M.S.Applied Artificial Intelligence
CISSP NVIDIA Generative AI and LLMs NVIDIA AI Infrastructure and Operations SIA AI Advisory Board Hands on CMMC Level 2 and SOC 2
Airman AI emblem
30 minutes / your documents

See it work on your own files.

Bring a few documents your team searches every week. I'll show you a private AI answering from them, live, on a box that never talks to the cloud.

Book the demo 907.947.2478