Vibe-coded MVP → production
Your AI‑built app has customers. Now it needs engineering..
You shipped on Lovable, Bolt, v0, Cursor or Replit and it worked, until security, data and scale became real questions. Golux Group audits AI-generated apps and rebuilds the parts that carry risk into software you own.
- Migrating from
- Lovable
- Bolt
- v0
- Cursor
- Replit
- Supabase
Why now
The prototype proved the idea. The next buyer will ask about everything else.
An enterprise customer sends a security questionnaire. An investor sends a due-diligence list. The hosting bill grows faster than usage. These are the questions a generated codebase cannot answer, and the reason founders call us. We answer them with a written audit before a line is rebuilt, and we rebuild onto a stack your future engineers can hire for.
- TypeScript
- React
- Ruby on Rails
- PostgreSQL
- Python
- LLM APIs
Client names and reference stories are shared on request, with their permission.
Where to start
One free door, and three ways to work together after
The audit and the first conversation are free and always will be. The work is priced for senior engineering. We take a small number of companies at a time, and doing that well is the whole business model.
- Free
Production-readiness audit
Paste your app's URL. In about a minute: exposed keys, tables readable without login, missing security headers, hosting you do not own.
Run the audit - From $2,500/mo
Fractional CTO
Six hours a month of senior technical judgment in writing, sixteen with a weekly call, or thirty-six embedded. The architecture, the hires, the answers for due diligence.
See the tiers - From $2,500/mo
Product design consulting
The product decisions before the pixels: what each screen is for, the flow that survives the edge cases, a design system your team and your AI tool can both follow.
See the tiers - From $20,000/mo
Startup Partner
Product, design, architecture and engineering as one arrangement, under one accountable person. One company at a time.
See the partnership
Looking for a discipline rather than an offer? AI, data, backend, web, mobile, design and growth engineering.
AI engineering spotlight
What it takes for an AI feature to survive production
AI engineering is the work of turning a promising model into a system your company can rely on: connected to your data, measured against real outcomes, and safe enough to put in front of customers and staff. It is less about the model and more about everything around it.
Grounded in your data
Answers drawn from your own documents and systems, with sources attached, so people can verify what they are told instead of trusting a confident paragraph.
Measured before launch
Evaluation against real examples with known answers, and quality, cost and latency thresholds agreed before anything reaches your users.
AI agents with boundaries
Agents that carry out multi-step work (gather, decide, act) inside clear limits, with an approval step wherever the risk sits.
See AI agentsWatched after launch
Quality, cost and behavior monitored continuously, so you know when something drifts before your customers tell you.
Engagement models
Buy a project, or buy judgment every month
Some problems have edges you can scope. Others need someone senior thinking with you every week. We do both, and we will tell you which one fits your situation.
Buy a project
A defined piece of work with a written scope: a fix, a migration, a feature or a first release. Project work is $300–$400 an hour with a 10 hours minimum, and the hours are agreed in writing after a call, before anything starts.
- Best for a named problem or a first release
- Scope and hours written down before work starts
- Documented handover, so nothing requires us to stay
Buy judgment every month
Senior technical judgment as a monthly arrangement: Advisory at $2,500 for 6 hours, Fractional CTO at $6,000 for 16 hours, Embedded at $12,000 for 36 hours. The hours are always stated, never unlimited.
- Best for a roadmap that needs senior decisions every week
- The same person on every call, and in writing between calls
- Move between tiers as the roadmap changes
Selected work
Products we built and still run
Every product below is live and operated by us, so you can open it and judge the work directly.
Monitoring SaaS
Server Inspector
Uptime, SSL, domain and API monitoring that tells you something broke before your customers do.
View projectAI matching platform
Work Transcendence
A knowledge graph of what a person knows, where every claim keeps the sentence that proves it, matched to real work requirement by requirement.
View projectEmail infrastructure
PigeonAtlas
Email sending for applications: an HTTP API and an SMTP endpoint, Resend-compatible, with your own DKIM key and newsletters that can never delay a password reset.
View projectAI marketing workspace
Merlin Studio
An AI marketing workspace where strategy, brand knowledge and content live in one place, and a person approves the final version.
View projectWellness SaaS
Ars Respirandi
A breathing practice anchored to one objective daily measurement, with an AI coach that reads the user's own data.
View projectAI product
LifeCopilot
A project execution system that turns a brief into sized tasks and tells you whether the deadline is real.
View projectDelivery process
Eight stages, from first conversation to continuous improvement
A process you can plan around: a demo every two weeks, written decisions, one point of contact, and the whole record in Golux Club.
- 01
Discovery and business alignment
The business goal, the users, the constraints and the data you already hold, and the metric success will be measured by.
- 02
AI and technical feasibility
Where AI genuinely earns its place, what stays conventional software, and what it costs to run, tested on your data before you commit.
- 03
Product definition
A first release small enough to ship and large enough to be useful, with everything deliberately left out written down.
- 04
Experience design and architecture
Interface and system design in parallel: prototypes people can react to, and an architecture that still holds at ten times the volume.
- 05
Iterative engineering
Short cycles with working software at the end of each one, a demo every two weeks and written decisions you can review.
- 06
AI evaluation and quality assurance
Automated tests, security review, performance work and, for AI features, evaluation against real examples before anything reaches your users.
- 07
Launch and integration
A controlled release into production, connected to the systems around it, with rollback tested rather than assumed.
- 08
Monitoring and continuous improvement
Quality, cost and usage watched in production, and a roadmap for the next improvement rather than a farewell email.
Why Golux Group
Six reasons companies keep the engagement running
AI and software under one roof
Models, product and platform engineering are decided by the same person, so an AI feature is shipped into a real system rather than demonstrated beside one.
Business-first engineering
Work starts from the metric you need to move (hours saved, conversion, time to decision, cost per transaction) and technology choices follow from it.
Senior, and named
The person you speak to on the first call is the person who does the work. Collaborators join by name, with a stated reason, after you have met them.
Flexible engagement
Change the arrangement, not the person: move between advisory, fractional and embedded hours as the roadmap changes.
Clear communication
One point of contact, a demo every two weeks and decisions written down. No status theater, no surprises at the end of a sprint.
Long-term product thinking
We build for the version of your product that exists in two years: documented, tested and maintainable by your own engineers.
About Golux Group
A founder-led practice, not a resourcing desk
Golux Group is led by Nikola Golux, who takes the first call, writes the architecture and, on most engagements, the code. Engineers and designers he has worked with for years join by name when a project needs them. Intelligent products fail far more often on unclear problems, weak data and poor experience than on model choice, so the person who scopes your work is the person who builds it.
More about usInsights
Practical writing on AI engineering and modern software
Guides on what breaks in AI-built apps, what it costs to fix, and how to choose who fixes it, written for the person who has to defend the decision.
Read the insightsStart the conversation
Start with the smallest version of your problem
Run the audit. It is free and it takes a minute. If one finding is the one that hurts, tell us and you will get a straight answer about what it is before anything is agreed. And if what you actually need is someone senior to think with every week, say that on the call; it is most of what we do.
