AI / Production
You built it
with AI.
Now ship it properly.
You shipped a working prototype with Lovable, Base44, Bolt, Cursor, or Claude Code. Getting it secure, stable, and live in front of real users is a different job. That last mile is mine: over a decade of production software, no rewrite, no judgement. Just shipped.
AI / Production / Capabilities
What's covered.
Security and secrets
AI scaffolds leak. Exposed API keys, endpoints with no auth, row-level security that was never switched on. I find and close the holes before your app goes anywhere near real users.
The missing spec
AI builds what you asked, not what you meant. I write the spec the prototype skipped, so the next feature does not quietly break the last one and you know what done actually means.
The production release
Hosting, environment config, a real database, CI, error tracking, a domain that stays up. The unglamorous last mile that turns a demo into something people can pay for.
Will it hold up
The vibe-coded version works for ten users. I find what breaks at ten thousand: the N+1 queries, the missing indexes, the request that falls over under real traffic, before your launch finds them.
10+ years senior delivery
Why me
Senior delivery you don't have to project-manage.
Most of my recent inbound is the same story. You built something real with an AI builder like Lovable, Base44, Bolt, or Cursor, it works on your laptop, and now you are staring at the gap between a working demo and something you can actually charge for. I have spent over a decade on the production side of that gap. I am not here to rewrite your app or tell you the AI did it wrong. I am here to get it shipped, secured, and stable, and to leave you able to keep building on it.
- Security and secrets audit before anything goes public
- The spec and architecture your prototype skipped
- A real production release: hosting, CI, error tracking, a domain that holds
- You keep ownership and can keep building after I hand it back
Engagement / Snapshot
A typical ship your ai app engagement.
Lead time
<7
Days to start, contract permitting
Length
4-12
Weeks for fixed-scope work
Reply
<24
Hours, every working day
Day rate
POA
Quoted per engagement
FAQ / Ship Your AI App
Frequently asked
Q · 01
I built my app with Lovable, Base44, Bolt, v0, Cursor, or Claude Code. Can you work with that?
Yes, that is exactly the work. Lovable, Base44, Bolt, v0, Replit, a raw Cursor or Claude Code project, it does not matter which tool generated the first version. I care that it ships securely and does not fall over. I pick it up where you are and take it to production.
Q · 02
Is my AI-generated code safe to launch?
Often not yet, and that is normal. The common issues are exposed keys, endpoints with no authentication, missing row-level security, and no input validation. None of it is hard to fix once you know where to look. A review tells you exactly what to close before launch.
Q · 03
What is a discovery day?
A focused paid day where I go through your app properly: security, architecture, what is missing for production, and what will break under load. You leave with a prioritised plan you can action with or without me. Most people book one after a free intro call.
Q · 04
Do I have to hand the whole thing over?
No. Plenty of engagements are a review plus a few days of hardening, then you carry on. I am happy to do the whole production push or just unblock the part you are stuck on.
Q · 05
How do we start?
A short free call. You show me the app, I tell you honestly whether I can help and what it would take. No pressure, no obligation.
How I work
Remote-first. Async by default.
Inspired by Remote and the way Basecamp ran a distributed team without burning anyone out. Meetings are expensive, writing is cheap, and momentum belongs to whoever can make a decision without a calendar invite.
01 / Specs
Specs before code
A short written spec covering problem, scope, and acceptance before anything gets built. Cuts rework, makes review easier, and means we never argue about what was agreed.
02 / Workstreams
Workstreams over standups
Active work lives in tickets and PRs with running notes. Async updates instead of daily standups, so deep work is the default and meetings only happen when they earn it.
03 / Loom > Zoom
Loom over meetings
Most reviews and walkthroughs go out as a 5-minute Loom you can watch at 2× when convenient. We jump on Meet / Zoom / Teams when the topic actually needs a conversation.
Channels / Tools
Slack
Async, threaded, no expectation of instant reply
Loom
PR walkthroughs, demos, design reviews
Google Meet / Zoom / Teams
Booked when conversation > writing
Linear / Jira / GitHub
Source of truth for the work
Notion / Docs
Specs, decision records, retros
The pitch
I write the code, talk to your team, and ship what we agreed. No agencies. No PMs. No surprises.
James Donnelly
Freelance developer · Manchester