Ship your AI product
like you ship code.
You shipped your MVP with v0, Cursor, Lovable. Users came. They aren't sticking. 92% of AI-built products have no empty states. 78% no error states. All have generic spinners. That's where retention dies — and that's what I fix. I shipped Lumen solo with Claude, run Vigilo in production across 191 countries, and turn AI-generated prototypes into products people actually keep using.
Vibe-coded products ship fast. Then they quietly bleed users.
Every AI-built MVP follows the same trajectory. v0/Cursor/Lovable get you to launch in days. By month 3, retention is broken — and the founder can't pinpoint why. The data is depressingly consistent.
No empty states
NN/g audit of 50 AI-built dashboards. New users open the app, see "no data" or a blank table, leave. Adding three states alone lifted retention by 12% in their tests.
No error states
API fails, user sees generic error or blank screen. No retry, no explanation. Each error is a churn event you don't see in analytics.
Generic UI
v0, Lovable, Bolt — all trained on the same shadcn/Tailwind/Vercel templates. Indigo accent, left sidebar, rounded cards. Investors and users can't tell you apart from 50 other AI tools.
Onboarding loses 60%
AI generates components, not flows. There's a sign-up screen and a main screen — nothing connecting them. By the time users figure out what to do, most have left.
I built this in a week. Solo. With Claude.
Lumen is a fully clickable AI video editor prototype — 21 screens, working onboarding, simulated AI generation, deployed on Vercel. No mockups. Real React, real interactions. Click around. This is what I ship.
Click "Get started" — explore Generate, Edit, Color, Sound, Deliver. Everything works.
Same outcome. 10× the velocity.
Traditional agency
- 2-week discovery phase before any pixels
- 5 people on Zoom every Tuesday
- High retainer minimums
- Figma handoff → 4 weeks of "implementation"
- Designs don't match what gets built
- You manage the agency, not your product
Stepikin Studio
- 30-min discovery call, day 1 we start
- One person. Direct DMs. No PMs.
- Flat rate. Fixed scope, fixed price.
- I ship the code. Live URL day 7.
- What you see is what ships, because I build it
- You ship the product. I'm invisible.
Three packages. Start small, prove value, then scale.
Start with a 5-day audit and see what's broken. If we agree on scope, move into a 4-week build. Fixed prices, no surprises.
AI Trust Audit
- One 45-min kick-off call and product access
- Expert workflow review: where users ignore, override or doubt the AI
- 10+ ranked fixes with impact and effort
- Your most critical flow redesigned in code
Live in 4 Weeks
- Discovery, UX and a scope locked on day 3
- Design system, UI and frontend in one pass
- Backend, auth, AI APIs and deploy
- A live build every Friday
- Verification report: tests and an AI eval set
- Code you own + 30 days of post-ship support
Ship Every Week
- Embedded design + engineering in your tools
- New features and AI surfaces from your roadmap
- Design system kept current
- Async-first: Loom + written updates
Four weeks. Four checkpoints. One person to talk to.
Strategy + flows
30-min kickoff. I map your product flows, ICP, and what to ship vs. cut. By day 3 you see a clickable Figma flow of the whole MVP.
Design system + UI
Build the design system in Figma (tokens, components, dark/light). Design all key screens. You get a Figma link with daily updates.
Build + integrate
I ship React/Next.js code as I design. Connect your AI APIs. Auth, billing, deploy. By day 21 you have a live URL.
Polish + launch
Animations, edge cases, analytics, SEO. You launch. I'm available for 2 weeks of post-launch fixes at no extra cost.
Real products I've designed, shipped, and ran in production.
Vigilo
Real-time global risk intelligence. 44 verified data feeds, 26 serverless functions, hourly refresh. Every signal traces back to its original source. SaaS, NGOs free.
EngiBoard
Native desktop SaaS shipped to production in 4 weeks. Tauri 2, Next.js, Supabase. 4 platforms, 44 releases, 98 commits. Real users, real revenue.
Lumen Studio
AI video editor prototype — 21 screens, onboarding flow, simulated generation. Built with Next.js, deployed to Vercel. The proof you clicked above.
AI Dent
AI diagnostic interface for dental clinics. Designed the core workflow for radiograph analysis. Used by 1,000+ clinics globally.
Answers founders actually ask.
We already shipped with v0/Cursor/Lovable. Why do we need you?
Because AI tools shipped the easy 70%. The remaining 30% — empty states, error states, onboarding flow, mobile responsive, real branding, accessible form patterns — is where your retention dies. I'm the person who reads your AI-generated codebase, identifies what's leaking users, and ships the fixes in production code. Without throwing away what works.
How do I know it's a design problem and not engineering?
Easy way to tell: open your product in incognito as a brand-new user. Count the screens where you see "no data" with no explanation. Count the loading states that are just generic spinners. Try it on mobile. Try breaking a network request. If you hit walls in 5 minutes, that's not engineering — that's design debt. The audit catches all of it on day one.
Why solo? Aren't agencies more reliable?
An agency reliably gives you 3 layers of project management between you and the work. I give you my direct phone. No briefs lost in translation. No PM bandwidth tax. If something's broken at 11pm, you ping me and I fix it. With AI tools (Claude, Cursor) one experienced person moves faster than a 5-person agency. The math has changed.
What stack do you ship?
React, Next.js, Tailwind, TypeScript. For backend: Supabase, Vercel, or your existing infra. For AI: OpenAI, Anthropic, OpenRouter, Replicate, fal.ai. For desktop: Tauri. I won't fight your stack — I work in yours.
How do payments work?
I invoice through easystaff — you can pay by SWIFT wire OR by corporate card (Brex, Ramp, Mercury, regular Visa/Mastercard). 50% upfront, 50% on delivery. Net 7 terms. Proper English invoice. Same easy flow as paying any SaaS — no special procurement process.
What if we need ongoing design after the sprint?
Two paths. Either we continue with Ship Every Week ($3,000/month, at least one shipped improvement a week, 3-month minimum), or I help you hire a full-time designer and onboard them on the design system I built.
How do I know you won't ghost mid-sprint?
I have a public portfolio with 7 cases, a live working demo (Lumen above), a Github (arablex), and 9+ years of working with product teams. Sprint contracts have clear milestones with partial refunds if I miss them. I haven't yet.
What's your background?
Ex-Head of Design at AI Dent (AI healthtech, raised $14M, 1000+ enterprise customers). Before that, lead designer at multiple Series A startups. Based in Georgia, work in EST/CET hours by default, flexible.
Stop blocking your engineers.
Start shipping design.
20-minute call. No pitch, no slides. We talk about your product and decide if a sprint makes sense. If not, I'll point you to someone who fits.
Book your callTypical reply within 4 hours · EST/CET friendly