I make AI actually work for your business.
An open-source model running on hardware you own. A model tuned to your job, so it stops needing a page of instructions to behave. The harness around it that makes the whole thing reliable. All of it measured on your real work, at a cost per task you can see.
£500/mo
You build it. I give you direction, review and the measurements.
02 / A build£2,500–£5,000
Fixed scope, fixed price. I build one thing and hand it over.
03 / Managed Stack£500/mo
Once it is running, I keep it current as the models change.
Not tool recommendations, and not a wrapper around somebody else’s API. Working systems you own, built by one person who does the work himself.
Work on it with me, every month.
Most AI advice is someone telling you what worked for them. This is your tasks, your numbers, and your hardware — measured, so you can see whether it is actually working rather than hoping.
£500per month
Cancel any time. No minimum term.
This is for you if
You are already paying for AI and cannot tell whether it is working
You are weighing fine-tuning, your own hardware, or moving off a vendor
You want the method, not a done-for-you black box
It is not if
You want someone to take the whole thing away and hand it back finished
You have not started yet — there is nothing to measure
The intro call comes first — it is free, and it is how we both find out whether this is worth your 2 sessions a month.
Or hire me to build it for you.
Three things I build, each for one price agreed up front. No hourly rate to watch, and nothing to renegotiate halfway through.
Local AI Deployment
£3,500fixed
An open-source model running on your own hardware or your private cloud instead of somebody else’s API. I pick the model by testing it on your real work, set up the serving, and measure what each task actually costs against the API bill you are paying now. You finish with something your team can run without me.
3–4 weeks
Custom Fine-Tune
£2,500fixed
A model trained on your task or your character, so it stops needing a long prompt to behave. I shape the dataset, run the tune on rented training infrastructure, and build the test harness that shows it beat the model you started with. The weights are yours.
3–4 weeks
Harness Build
£5,000fixed
The research or agent harness around the model — the part that decides whether any of it is reliable. When the membership finds something worth building, I build it. Not wireframes and handoffs: working software, deployed and running. One person, full stack, start to finish.
4–6 weeks
Two build slots a month — when they’re taken, they’re taken.
How a build runs
Scope, and a look at your data
We agree exactly what gets built and what it is measured on. For a fine-tune I check the state of your data before you sign — free, and written into the agreement — because a fixed price should never quietly turn into a rescue job.
The clock starts when access arrives
Not at signature. Signing books the slot; the delivery clock starts the day the data, accounts and keys we need are actually in hand, and I confirm that date to you in writing.
The build, with the benchmark deciding when it is done
“Done” is not my opinion. It is the benchmark or eval we named at the start, run on your real tasks — so there is one agreed test that either passes or does not.
A demo and the numbers, whatever they say
You see the finished thing working — a walkthrough, a live run, and the benchmark results whatever they say. From there you have seven days to check it against the benchmark we agreed at the start.
Handover on acceptance, then 30 days of fixes
Everything transfers to you once you accept it: weights, code, configuration, the handover doc. Nothing is held back pending payment, because it was paid at the start. Defects in what was delivered are fixed free for 30 days.
The terms
Payment in full books the slot
Paid upfront — with two slots a month, that is what takes yours off the market. Cancel before we start and you get all of it back; cancel after, and I refund the part I have not earned.
You see it working before it is yours
I show you the finished work and the benchmark, whatever it says. You have seven days to check it against what we agreed, and everything transfers to you on acceptance — for a fine-tune, the weights included.
Thirty days of fixes, free
Defects in what was delivered — things that do not do what we agreed they do. New models, new scope and changes on your side are not defects.
Then it is Managed Stack at £500/mo, or a new engagement
There is no third door where support quietly continues for free. A build ends with a handover, not a dependency.
What I do not sell
—No hourly consulting
—No training infrastructure hosted for you
—No hardware supplied
—No open-ended retainers
A build ends with a handover, not a dependency. If you want it kept current after that, Managed Stack is the way — it is offered at the end of every build, never assumed.
Managed Stack
Your stack, kept current: new models evaluated against your tasks and pushed to you; repricings, deprecations and terms changes caught before they cost you.
£500per month
Cancel any time. No minimum term.
One person built all of this.
MenFem is a live platform — not a demo. Every system below was designed, coded, deployed, and is actively maintained by a single operator using AI-native workflows.
Plus: real-time SSE, background workers, image pipeline, Redis caching, Stripe Connect payments, email system, admin dashboard — all one codebase, one person.
Book a 20-minute diagnostic call.
No pitch, no deck. If it’s a fit, I’ll send a one-pager. If it’s not, I’ll tell you what I’d do instead. I respond within 48 hours.