I’m an independent engineer. I work across Azure, LLM integration, and the data plumbing in between. Teams usually call me when the infrastructure has outgrown whatever plan it was built on, or when an AI idea has to turn into something customers can actually use.
Services
Every engagement is scoped and priced in writing before it starts. No hourly meter, and no open-ended retainer you’ll forget to cancel.
The bill climbs faster than the traffic. Deploys still happen by hand. And nobody can say with any confidence what’s actually running in the subscription.
You’ve got real users and a long list of AI ideas. What you don’t have is a route from a convincing demo to something you’d put in front of a paying customer.
Two systems that ought to share data don’t. So every week somebody exports a spreadsheet out of one and re-types it into the other.
Scope
Saying this up front saves us both a call. If your problem is on this list, I’m the wrong person, and you should find that out now rather than three weeks in.
Open-ended staff augmentation with no defined outcome. If the brief is “a developer for six months”, you want an agency bench, not me.
A half-day task costs more in context-switching than it delivers. I’ll tell you that instead of invoicing for it.
Where verifying an answer costs as much as producing one and mistakes are expensive, automation makes things worse. I’ll say so even if the budget is already signed off.
A bad form, an unclear policy, a process nobody owns. Software built on top of those gets expensive and stays broken. Fix the cause first, and you may not need me at all.
If a job needs depth I don’t have, I’ll say so and point you at someone who does. That happens a few times a year and it has never once cost me a client worth having.
Fixing a system I can’t read, deploy or test isn’t consulting, it’s guessing. Repository access and somewhere safe to break things, or it isn’t a project.
The question worth asking first
Not by being impressive in a demo. Cheaper, faster cognition turns into money through six mechanisms and no others. Most disappointing AI projects either picked the wrong one, or leaked the value before it ever reached the P&L.
The machine does work a person used to do. It’s the route everyone talks about, and the hardest one to actually land. It pays only when checking an answer is far cheaper than producing one, and when the freed-up capacity genuinely gets removed or redeployed.
The same people produce more, or better. It’s where most real value sits today. It’s also the route people over-claim most, because minutes saved across thirty staff never turn into a budget line by themselves.
Work nobody did, because it cost too much, suddenly becomes economic. Reading every support call instead of sampling two percent. Reviewing every contract, not the ten someone flagged. Usually the biggest mechanism, and the most overlooked, because there’s no existing cost line to compare it against.
The same output, much sooner. Wherever latency carries a price (lead response, incident resolution, time to quote) compressing it is worth money on its own, whatever happens to labour. Easy to measure, and easy to attribute.
The rule gets applied the same way at 9am Monday and 5pm Friday. Sometimes the variance is the cost: compliance exposure, uneven pricing, audit findings. Removing it is the whole value, even if total labour doesn’t move.
Capability you can sell. Natural-language access to your own data, analysis as a paid tier, personalisation at scale. This one shows up as revenue, not cost, and it’s the only mechanism where being early is worth anything by itself.
One question predicts which of these you’ll get: how much cheaper is checking an answer than producing one? Where verification is cheap, automation compounds. Where checking costs about as much as doing the work, you get augmentation at best, and that changes the business case completely.
How it works
Thirty minutes, no charge. What’s broken, what finished looks like, and whether I’m the right person for it.
Fixed scope, named deliverables, a timeline and a price. Nothing starts until you’ve approved it.
Code lands in your repository from day one, and there’s a demo every week. No black box, and no nasty surprise at the end.
Documentation, a runbook, and a walkthrough with your team, so the work doesn’t depend on me sticking around.
Questions
It depends on the size of the estate, which is why I quote per engagement instead of publishing a rate card. A single-subscription setup is usually a week or two of work. A multi-subscription estate carrying legacy workloads takes longer. Either way, you get a fixed price in the proposal before anything starts.
Most of the time, yes. I work as an extra pair of hands inside your repositories and your ticket tracker, not as an outside supplier lobbing deliverables over a wall. Where a team already has strong opinions about tooling, I work within them instead of relitigating the decision.
That gets decided in writing before any code is written: which data leaves your systems, which model provider processes it, what’s retained and for how long. Where the data is sensitive, the options include regional endpoints, zero-retention terms with the provider, and keeping the sensitive fields out of prompts altogether.
Yes. Send yours over and I’ll sign it, or I can provide a short mutual one. Much better to do that before a discovery call than halfway through a project.
Yes, within reason. A focused integration or a week-long review is a good fit. Anything under a couple of days usually costs more in context-switching than it delivers, and I’ll tell you so instead of taking the work.
You do. Ownership of everything produced for an engagement transfers to you, and it lives in your repository from the first commit. I keep no private copies, and I reuse nothing specific to your business.
Mostly European and North American teams. I keep several hours of daily overlap with whichever of those you’re in, and meetings get scheduled inside your working day, not mine.
Handover includes documentation and a walkthrough, so your team can maintain the work without me. If you want me to stay involved, a small monthly retainer covers questions, dependency updates and occasional changes. It’s optional, and there’s no lock-in.
Proof
Because I’d have to either invent them or breach an NDA. Most of the work sits behind confidentiality agreements, and the rest isn’t mine to advertise.
The usual fix is stock testimonials, a “leading fintech” case study with the details filed off, and metrics nobody can verify. I’d rather you checked the thinking directly, which you can do before speaking to me at all:
Contact
Tell me what you’re working on and what’s in the way. If I’m not the right fit I’ll say so, and point you somewhere better.
One business day, usually sooner. Every enquiry gets a reply, including the ones I turn down.
A sentence on the system involved, what you’ve already tried, and any deadline you’re working to. Enough for me to tell you honestly whether I can help.
Happy to sign an NDA before we talk specifics. What you send is only ever used to reply to you, and it’s never sold.