Fix your AI-generated code before it breaks for real users.

Getting a working app out of an AI tool is the easy half now. The hard half shows up later: a login check that only happens in the browser, an API key shipped to every visitor, a query that's fine with ten users and falls over at a thousand, and a codebase nobody on the team fully understands. I read the code the AI wrote, find what's actually dangerous, fix it, and leave you with something your next developer can work on.

Tell me about the job

Booking · Sep '26

Job card · 01sajawal.dev

Fix AI-generated code

Sounds like you if

The app works in the demo, but you get nervous every time someone new signs up
It was built with Cursor, Lovable, Bolt, Replit, v0 or Claude Code, and nobody has read all of it
Fixing one bug makes two more, and the AI keeps suggesting the same broken fix
You're about to raise, sign a big customer or go through a security review
Your hosting or AI bill is growing faster than your users

Tools on the bench

Next.js
React
TypeScript
Python
FastAPI
PostgreSQL
Vercel
AWS
Azure

How I’d handle it

The job, step by step.

01
Read all of it
The code, the database and the deploy setup, plus the paths a real user and a bored attacker would take. You get a written list of what's wrong, ranked by how much it can hurt you.
02
Fix what can hurt you
Security and data problems first: exposed keys, missing auth checks, open database rules, unchecked input. Then whatever breaks under load. Cosmetic rewrites wait, or never happen.
03
Make it safe to change
Tests around the flows that make you money, error tracking, a proper deploy pipeline and short notes on how it fits together, so the next change, by a person or an AI, doesn't undo the work.

What you walk away with.

01A written audit: every issue, where it lives, how bad it is and what it takes to fix

02Security fixes for auth, secrets, database access and input handling

03Performance fixes for the queries and AI calls that are slow or expensive

04Tests and error tracking on the flows that matter

05A handover call and notes, so your team or your next hire can carry on

Shipped, not just pitched.

Products of mine where the same work is running for real users.

Placewise AILiveVoice agents that interview and screen candidates on autopilot.70% less manual screening

ReachWise AILiveCustom AI agents for businesses: voice, chat and automation that actually ships.90% automation rate

DocketryIn betaTill, kitchen screens, stock and delivery for UK fast food, in one system.Pilot shops onboarding

Questions people ask first.

Which AI coding tools do you work with?

Code from any of them. Cursor, Claude Code, Lovable, Bolt, Replit, v0, Copilot: once it's in your repo it's just code, and that's what I read. I use AI tools in my own work every day, so I know the mistakes they tend to make.

Do I need to rewrite the whole thing?

Almost never. Most AI-built apps have a sound idea and a handful of serious problems, and fixing those is faster and cheaper than starting again. If a rewrite really is the cheaper path, the audit will say so and show why.

How long does it take, and what does it cost?

It depends on the size of the app. The audit comes first, and it ends with a fixed quote for the fixes, so you know the cost before any of that work starts.

Can you work with my existing developer?

Yes. I can make the fixes myself, pair with your developer, or review their changes as they go. A lot of teams want the second pair of eyes more than extra hands.

My stack isn't on your list. Can you still help?

Maybe. Tell me what it's built on in the contact form and I'll say straight away whether I'm the right person, and point you to someone better if I'm not.