What qualifies as a rescue project?

Anything that is shipped but broken, inherited and unreadable, or built by a previous agency and now unmaintainable. If you are afraid to deploy on Fridays, that qualifies. I have taken over projects in Rails, Django, PHP, and every Node flavour.

How long does a rescue engagement take?

It depends on the codebase. The first week is always a paid audit - I read everything and give you a written triage report. From there, rescue engagements typically run 2-6 weeks. We agree scope before any code changes.

Will you rewrite everything from scratch?

Rarely. A rewrite is usually slower and riskier than a targeted refactor. I keep what works, replace what does not, and document everything so your team is not dependent on me going forward.

Can you work alongside my existing team?

Yes - often preferred. I can review PRs, pair on the hardest parts, and upskill your team on the patterns I introduce. You get a fixed result and your team gets stronger. Just give me access to the repo and a Slack channel.

AI PROTOTYPE RESCUE

Your prototype works. Your launch doesn't.

Your AI-built prototype is impressive. Let's make it production-safe before your next demo breaks it.

Get a free codebase assessment →from $4,500

SOUND FAMILIAR?

If you're nodding at any of these, we should talk.

✓"My Lovable app crashes when two users log in at the same time."
✓"Authentication is a hardcoded user — I cannot add real sign-up without breaking everything."
✓"Stripe works in test mode. I have never tested a real webhook and I am scared to."
✓"The database is SQLite and I do not know how to migrate it without losing production data."
✓"The AI wrote 4,000 lines in three sessions. I cannot explain half of what it does."
✓"My investor wants to see the product live next week and it crashes on mobile Safari."
✓"I have three different authentication systems in the same codebase and none of them are complete."

OVERVIEW

Lovable, Cursor, v0, and Bolt can produce a convincing prototype in an afternoon. What they cannot produce is a system that handles 50 concurrent users, processes real Stripe webhooks without data loss, or stores customer information without exposing it. If you have an AI-generated codebase that works in your screen recording but fails when someone else tries it — that is exactly where this engagement starts. Week one is a written codebase audit you receive before we scope the rest. You decide what to keep, what to replace, and what the budget looks like. I refactor only what will cause a real production failure. No rewrites for vanity.

THE RESCUE PLAN

Let's see what's actually broken.

Codebase Audit

WEEK 1

Week one: I read every file, document security risks, architecture gaps, and silent failure points. You receive a written report before we scope the rescue work. You decide what to fix and what budget makes sense.

Foundation Replacement

WEEKS 2–3

Swap out the unsafe parts: real authentication (Clerk or NextAuth), a real database (Postgres/Supabase), proper environment variables, secrets management, and a CI/CD pipeline that does not deploy broken code.

Feature Stabilisation

WEEKS 3–5

Reconnect the existing UI to the new backend. Add error boundaries, fix webhook handlers (Stripe, email, third-party APIs), and replace silent failures with observable ones.

Handover

WEEK 5–6

End-to-end test pass, security review, runbook, architectural decision record, and a 30-day support window. The next engineer gets documentation they can actually use.

IDEAL FOR

Right for you if these ring true.

→You built a prototype with Lovable, Cursor, v0, Bolt, or similar AI coding tools and it actually works — in demo conditions.
→You have a pilot date, an investor meeting, or real users lined up and the codebase is not ready for them.
→You want an honest assessment of what is safe to keep before committing to more development spend.
→You need real authentication (not hardcoded users), a real database (not a JSON file), and webhooks that do not silently fail.
→You have been told "just rewrite the whole thing" and you suspect that is not actually necessary.

OUT OF SCOPE

What I don't do.

✗I do not rewrite working UI that is not a security risk — if it renders correctly, it stays.
✗I do not expand the product scope beyond what already exists in the prototype.
✗I do not take responsibility for features the AI generated that were never demoed or tested.

OUT OF SCOPE

Not right for this?

✗Prototypes that have not been shown to any potential user — run a discovery sprint first.
✗Codebases where the entire architecture needs replacing (a full MVP Build is faster and cheaper).
✗Teams that want to preserve every line of AI-generated code — some of it has to go.

- FOR THE RECORD

The AI Prototype Rescue service by Suhag Al Amin converts AI-generated prototypes built with Lovable, Cursor, v0, or Bolt into production-ready products in 3–6 weeks. The engagement starts with a written codebase audit delivered in week one — the founder reviews the findings and decides what to keep and what to replace before any additional spend is committed. Suhag then installs production authentication, a real database (Postgres or Supabase), CI/CD pipelines, and reliable webhook handling. Only code that causes production failures is replaced. Pricing starts from $4,500 USD. Inquiries: suhag.alamin13@gmail.com or https://cal.com/suhag.

Have a pilot deadline? Let's talk.

Tell me where you are. I'll tell you, honestly, whether 6-8 weeks is realistic and what the first week looks like.

Book a 30-min intro call →suhag.alamin13@gmail.com