Your AI prototype, live and taking payments by Sevastian RakhimovYour AI prototype, live and taking payments by Sevastian Rakhimov
Your AI prototype, live and taking paymentsSevastian Rakhimov

Who this is for

You have something that works. A vibe-coded app, an AI demo, an MVP an agency left you with, a weekend project that unexpectedly got users. It does the impressive part well. It is also the reason you cannot sleep: there is no real auth, errors show a blank screen, nobody knows what it costs to run, and you have no idea what happens when a hundred people use it at once.
You do not need it rewritten. You need it to survive contact with real users and take money.

What I do

I take the thing you already have and close the gap between "it works when I show it" and "it works when I am asleep".
That usually means, in this order:
Read it and tell you the truth. A day of going through the code, the data model and the infrastructure, and a written list of what will break first, ranked. You get this before anything else, and you can stop here if you want.
Make it safe to have users. Real authentication and authorisation, secrets out of the source code, input validation, the obvious ways in closed.
Make failure visible instead of silent. Structured error handling, logging you can search, monitoring and alerts that reach a human. For AI features specifically: evaluation on a reference set, tracing of every call, and cost ceilings — an LLM feature nobody measures degrades quietly.
Make it deployable by someone who is not me. CI/CD, environments, migrations, a rollback that has actually been tested.
Take payments if it needs to. Subscriptions or one-off, and the boring parts that decide whether the money arrives — webhooks, restores, reconciliation.

Why me

I did exactly this on my own product. F/AI is an AI fitness trainer I co-founded and still run as CTO: 340,000+ registered users, 23,000 monthly actives, $8,000 MRR and 2,000+ paid subscriptions, live in the App Store and Google Play. The user counter is public at fitgpt.pro/stats if you want to check it rather than believe it.
Before that, six years with a New York SaaS company — the first eighteen months of it as the only engineer.
I have been the person alone with a product that has to work. That is the situation you are in.

How we work

Fixed price, agreed after I have seen the code — not before. Written scope, so we both know what is in and what is out. You get commits as they land, not a reveal at the end. I am in Frankfurt, Germany (CET), work in English, and take calls without a translator.
FAQs

Starting at$4,700
Tags
Flutter
PostgreSQL
Python
Supabase
AI Agent Developer
AI Engineer
Backend Engineer
DevOps Engineer
LLM
Mobile Engineer
Service provided by
Sevastian Rakhimov Tbilisi, Georgia
2
Followers
Your AI prototype, live and taking paymentsSevastian Rakhimov
Starting at$4,700
Tags
Flutter
PostgreSQL
Python
Supabase
AI Agent Developer
AI Engineer
Backend Engineer
DevOps Engineer
LLM
Mobile Engineer

Who this is for

You have something that works. A vibe-coded app, an AI demo, an MVP an agency left you with, a weekend project that unexpectedly got users. It does the impressive part well. It is also the reason you cannot sleep: there is no real auth, errors show a blank screen, nobody knows what it costs to run, and you have no idea what happens when a hundred people use it at once.
You do not need it rewritten. You need it to survive contact with real users and take money.

What I do

I take the thing you already have and close the gap between "it works when I show it" and "it works when I am asleep".
That usually means, in this order:
Read it and tell you the truth. A day of going through the code, the data model and the infrastructure, and a written list of what will break first, ranked. You get this before anything else, and you can stop here if you want.
Make it safe to have users. Real authentication and authorisation, secrets out of the source code, input validation, the obvious ways in closed.
Make failure visible instead of silent. Structured error handling, logging you can search, monitoring and alerts that reach a human. For AI features specifically: evaluation on a reference set, tracing of every call, and cost ceilings — an LLM feature nobody measures degrades quietly.
Make it deployable by someone who is not me. CI/CD, environments, migrations, a rollback that has actually been tested.
Take payments if it needs to. Subscriptions or one-off, and the boring parts that decide whether the money arrives — webhooks, restores, reconciliation.

Why me

I did exactly this on my own product. F/AI is an AI fitness trainer I co-founded and still run as CTO: 340,000+ registered users, 23,000 monthly actives, $8,000 MRR and 2,000+ paid subscriptions, live in the App Store and Google Play. The user counter is public at fitgpt.pro/stats if you want to check it rather than believe it.
Before that, six years with a New York SaaS company — the first eighteen months of it as the only engineer.
I have been the person alone with a product that has to work. That is the situation you are in.

How we work

Fixed price, agreed after I have seen the code — not before. Written scope, so we both know what is in and what is out. You get commits as they land, not a reveal at the end. I am in Frankfurt, Germany (CET), work in English, and take calls without a translator.
FAQs

$4,700