AI consulting & development · Las Vegas + remote
AI help from the person who actually builds it.
AI can feel overwhelming, and most of the help out there is either a strategy deck or a monthly bill that keeps growing. I’ve been building software for thirty years. I’ll explain your options in plain English, then build and host the custom AI, automation and integrations myself, on hardware I own.
01 / The problem with most AI help
You don’t need another strategy deck.
You need something that works. Here’s what usually gets in the way.
Handed off to someone else
An agency scopes it, then a junior builds it, and you never talk to the person writing the code.
A bill that grows with every use
The “AI solution” turns out to be a thin layer over someone else’s service, metered per request.
Your data leaves the building
To make it work, your private information gets sent to a third party’s servers.
Meetings instead of software
Weeks of discovery calls later, there’s still nothing working you can show a customer.
02 / Who you’ll work with
Hi, I’m Phil.
I’ve been building software for thirty years, and I do this myself, with my assistant. The person you describe the problem to is the person who designs it, builds it and keeps it running. Whether you need an hour of straight answers or a whole system built, you’re talking to the builder.
03 / Why work with me
Senior engineering, without the agency layers.
Talk to the builder
The person who picks up the phone is the person building it. No account managers, no hand-offs.
Own your stack
Models can run on hardware I own, with no per-request meter that punishes you for growing.
Private when it matters
Sensitive information can be processed entirely on equipment I control.
Thirty years of shipping
From a Lake Tahoe garage to a thousand servers in San Francisco to an AI studio in Las Vegas.
Working software, not slides
You get something you can put in front of customers, not a strategy deck.
Frontier AI when it fits
GPT or Claude for the hardest problems, local models for privacy and cost. The right tool for each job.
04 / My own AI hardware
I don’t rent AI. I run it.
Most “AI companies” are a login to somebody else’s service. I run a real cluster of my own: 13 machines and about 2.2 TB of memory, enough to run a 671-billion-parameter model locally. That means private by default, and no meter running on every request.
Mac Studio · M3 Ultra
Up to 512 GB each, linked into a cluster that runs the largest open models end to end.
NVIDIA DGX Spark
Grace-Blackwell machines for image and video generation and GPU work.
Minisforum MS-S1
Always-on AI, a private document archive, and search over your documents.
Mac Mini
Small machines running automations and agents around the clock.
05 / The honest math
Does AI cost money? Yes, quite a bit.
Here’s my own AI stack, every month, on autopay. It all runs alongside about $53,400 of AI hardware in my office.
| ChatGPT Pro · frontier model | $200.00 |
| Claude · frontier model | $200.00 |
| OpenRouter · model credits | ~$150.00 |
| Higgsfield · video and images | $49.00 |
| Perplexity · research | $40.00 |
| Lovable · app builder | $25.00 |
| Firecrawl · web data | $19.00 |
| Monthly total | ≈ $683 / mo |
You don’t pay for any of that. You pay $125 an hour for the work, and the tools and hardware are my problem, not yours.
06 / How it works
From “here’s the problem” to working software.
01
A free first conversation
Tell me what you’re trying to do. You’ll get straight answers, not a sales funnel.
02
A clear plan
What runs locally, what uses frontier AI, roughly how long, and what it will cost.
03
I build it
Piece by piece. You see real, working progress instead of status meetings.
04
Run it or hand it off
I can host and look after it on my own infrastructure, or hand it over cleanly. Your call.
07 / Tell me what you’re building
Request a free consultation.
Or reach me directly
Text me Call (725) 227-5702 oakweb@gmail.comWHAT HAPPENS NEXT
I read every request myself and reply, usually the same day, to set up a quick call.
08 / Straight answers
Common questions.
Do you actually build, or just advise?
Both. I take consulting-only work when you just need direction, and I build and ship full systems when you need it done. Same person either way.
Can I still use OpenAI or Claude?
Of course. I build hybrid systems: local open models for anything private or high-volume, and frontier AI like GPT or Claude for the hardest problems.
Is my data actually private?
When it needs to be, yes. I can run AI on hardware I own, so private or regulated information never has to leave for a third party. We decide together what runs where.
What does it cost?
The first conversation is free. After that it’s $125 an hour, with calls billed in 15-minute blocks and building billed piece by piece once it’s working. I’ll give you a realistic picture before any work starts. See how billing works →