I design the interface and write the software.
Technical cofounder of Aithority (EU AI Act) — backend, integrations, and classification logic; the company is in Lanzadera, Spain. F1 Strategy Agent open-sourced after viewers asked for the code, now extended into a real SQL data pipeline. Founder of Kiblo, in App Store review for the US and Canada. Dross notarized Mac. BlockFlow’s voice agent is in production, not sold like Kiblo.
TypeScript · Node · SQL · Next.js · React Native · SwiftUI. I use AI in loops: conditions first, then rebuild what failed — not one long prompt. ~2 years in Paderborn at Deutsche Post / DHL; based in Spain, open to backend/software engineer roles in the Netherlands and Germany.
Loop engineering. Lock the conditions. Own the product.
I do not send the model one long prompt and hope the result is right. I run loops with AI: build against explicit conditions; if they fail, inspect why, then rebuild. I own the constraints and the ship. The model does not.
Loops, not a lucky prompt
Each loop has conditions that must hold (tests, a visual lock, a rule that must not be faked). If they are not met, the next pass is “why did this fail?” then a correct build — not a longer prompt. Claude, Cursor, or whatever the job uses: same loop.
Problems first, then the stack
Kiblo: no LLM for feeding advice; RER for grams; real US recipes in the plan; home stays a bowl. Aithority: seven-minute setup down to a three-step confirm flow; Turso when Render’s disk wiped departments. Dross: compiler API, not regex.
Team outcomes, not a founder pitch
Solo shipping is proof I can carry a surface. On Aithority the commercial cofounder owns sales and legal; I own product and engineering. The job I want is the same split on a team: design and code, same person, shipped software.
What I bring, and what I’ve shipped.
Skills
Languages
TypeScript · Python · Dart · SQL · Swift
Frameworks
Next.js · React · React Native / Expo · Flutter · Node / Express · SwiftUI
Data & infra
Firebase · Turso / SQLite · Supabase · Vercel · Render · RevenueCat
Integrations
Stripe Connect · Microsoft Graph · Meta Ads API · OAuth · OpenAI / Anthropic admin APIs · Resend
AI
Claude API · LLM orchestration · Eval design · LLM-as-judge · Voice AI · Deterministic rules over LLM · RAG
Abilities
Own the backend and the integrations
Classification logic, admin-API connections, auth, data models, and the move off a platform when it fights the product (Render → Turso on Aithority).
Run AI in loops against locked constraints
Conditions first — tests, a visual lock, a rule that must not be faked — then rebuild what failed. I own the constraints and the ship; the model does not.
Ship to the App Store solo
IAP, subscriptions, privacy manifests, DAC7, UGC moderation, and Apple review — carried end to end on TRACE, SMASH and Kiblo.
Design the product and its visual system — no handoff
Interface, identity, typography and motion, then the code that ships it. I do not need a designer to put a product in front of users.
Distribute what I build
Short-form video, written, recorded and published solo. The F1 repo went public because viewers asked for the code.
Milestones
Aug 2026
TRACE live on the App Store (US)
Privacy app — breach scan, exposure score, broker removals. Shipped and public.
Aug 2026
Kiblo submitted — in App Store review (US & Canada)
Concept to review build in days, Cursor + Claude in loops.
2026
Dross shipped — notarized Mac app + CLI
Contract-drift checker with a golden corpus at 100% precision and recall.
2026
Aithority accepted into Lanzadera, first paying client
EU AI Act compliance. The company is in the accelerator — technical cofounder.
2024–2026
~2 years at Deutsche Post / DHL, Paderborn
Building and shipping products end to end alongside the job.
Live
SMASH on the App Store · Volea and Louvr Labs in production
Earlier products still running with real users and clients.