A production localization pipeline built entirely on the OpenAI API: one English source of truth, SEO-aware translation prompts, cross-model verification with
Iurii Fixes: Rescue Projects
Legacy modernisation for projects that are stuck, broken, behind schedule — or an AI-generated codebase that won't quite ship. I'll assess it honestly and get things moving, without rewriting from scratch.
Based in Finland (UTC+2) · VAT-registered FI29845875 · Respond within 24 hours
Recent client work
“
We took over a Next.js project from another agency and inherited a mess — failing builds, no tests, deployments that worked locally and broke in production. Iurii spent two weeks getting it stable before touching any new features. He documented what was broken, fixed the critical issues, and left us a short list of the smaller things we could handle ourselves. By the end the project was actually maintainable, which had felt impossible at the start.
Niamh O'Sullivan 🇮🇪
Agency Owner
You hired a developer, things went wrong, and now you have a codebase you can't ship — or a live system that's falling apart. Or your own team got overwhelmed and the project stalled six months ago.
There's a newer flavour of this too: codebases built mostly by AI — Cursor on autopilot, an LLM-driven weekend MVP, a junior dev who accepted every Copilot suggestion. The code compiles, the tests pass, the structure looks vaguely right. It's still broken — just in less obvious ways.
I've taken over projects in bad shape before, including both kinds. Most of this work is legacy application modernisation done surgically: the first thing I do is understand what's actually there before recommending what to do about it. Sometimes the code is salvageable with targeted refactoring. Sometimes a partial rewrite is necessary. I'll tell you the truth — including if the problem is something I can't solve.
This is not a service for finding someone to blame. It's for getting something working.
When the codebase was vibe-coded
AI-generated codebases have their own pathology. They look professional at a glance and fall apart on second read. The patterns are consistent enough that I now check for them first:
- Five overlapping ways to do the same thing — auth handled in middleware, pages, API routes, and a client hook, because the model never deleted the previous attempt
- Tests that pass because they assert what the code does, not what it should do — full green CI on broken behaviour
- Bloated dependency lists: every library the model has heard of, half of them unused, two of them deprecated
- Security holes copy-pasted from public examples — tokens in localStorage, wide-open CORS, SQL built by string concatenation
- Database migrations that don't reverse, schemas that drift from the code, foreign keys that exist on one side and not the other
- A README that confidently describes features that don't exist — written by the same model that built the app
What's included
- Full codebase audit: architecture, security, performance
- AI-generated codebases: dead-code removal, dependency cleanup, real test coverage
- Honest written assessment of what's there and what it'll take to fix
- Prioritised fix plan: critical bugs first, then tech debt
- Bug fixing, refactoring, performance improvements
- Documentation and knowledge transfer
How I approach this
01
Diagnosis first
I read the code, run the tests (if any), and map out what's broken and why. No assumptions.
02
Written assessment
A clear report: what the codebase looks like, what the risks are, what I recommend, and roughly what it'll cost to fix.
03
Agree on the plan
We prioritise together. Critical fixes first. You decide how far to go.
04
Systematic improvement
Work in short cycles with visible progress. No disappearing for three weeks and coming back with a surprise.
Relevant projects
View all projects
Internal sales portal that turns a wholesale deal into a full set of trade paperwork — pro forma, contract, commercial invoice, packing list, CMR and more —
Related articles
View all articles· 14 min
Vibe-coded codebase problems: duplicate auth flows, tests that assert nothing, unused dependencies, security shortcuts, schema drift — why they happen and how
· 16 min
AI coding widens the gap between senior and junior developers — it amplifies existing skill.
More Rescue Projects reviews
“
Our app had slowed to a crawl and the previous developer had left no documentation. Customers were starting to churn over load times. Iurii profiled it instead of guessing, found a handful of unindexed queries and a runaway N+1 that were doing most of the damage, and had page loads back under a second within the first week. He fixed the actual causes rather than throwing more server at it, and wrote up what he changed so we wouldn't repeat the same mistakes.
Owen Pritchard 🇬🇧
CTO
Databases
“
I'd built most of our MVP with Cursor and it looked finished — it compiled, the tests were green, the demo worked. It just wouldn't survive real users. Iurii took it over and found what I couldn't see: authentication handled four different ways, tests that only asserted what the code already did, and a dependency list half of which was unused. He didn't rewrite it from scratch — he told me honestly what was salvageable, ripped out the dead code, and got it to something a real team could build on. Two weeks and it went from 'looks done' to actually shippable.
Sebastian Falk 🇸🇪
Founder
AvailableReady to discuss your rescue projects project?
Tell me what you're building or stuck with. I'll reply within 24 hours.