I build at the point where product, AI, code, and visual storytelling meet.
AI Product Builder & Creative Technologist
I turn rough ideas into mobile products, AI systems, generative video, and things that leave the screen as physical objects.
Across layers
Tools I work with
CAPABILITIES
Seven areas, and the real tools behind each one.
Product decisions before a line of code exists
Deciding what a feature actually needs to do, where the free plan ends, and whether something's really ready to ship — the calls that shape a product long before there's a UI to look at.
- Feature architecture
- Onboarding thinking
- Monetization boundaries
- Release readiness
- Iterative feedback
Mobile apps that actually hold up
Building the real thing in Expo and React Native — navigation that doesn't fight you, OTA updates so fixes don't wait on a store review, and a Supabase backend underneath it all.
- Expo / React Native
- TypeScript
- Navigation
- OTA updates
- Supabase
- Widgets
AI integrations that don't fall over
Wiring Claude and Gemini into a product properly: structured outputs instead of parsing prose, a fallback model when the primary one is down, and API keys that never touch the client.
- Claude
- Gemini
- Structured outputs
- Fallback logic
- Server-side proxy
- Vision & voice
AI video, cut like a real production
Generating footage with Veo and Kling, but treating it like a real shoot — storyboards, shot continuity, and a Remotion edit at the end, not just stitched-together clips.
- Veo / Kling
- Storyboards
- Continuity
- Shot design
- Remotion
- AI music workflows
The backend glue that runs on its own
Small Supabase Edge Functions and Cloudflare Workers that handle webhooks, notifications, and Telegram bots without needing me to be there.
- Supabase Edge Functions
- Cloudflare Workers
- Webhooks
- Telegram bots
Shipping it myself, start to finish
I run my own Hetzner server through Coolify, so pushing to GitHub is what puts a change in front of real users — no separate ops person, no waiting on someone else.
- GitHub
- Coolify
- Hetzner
- Cloudflare
- Push-to-deploy
3D work that ends as a physical object
Modeling in Blender, Cinema 4D, and 3ds Max, then actually printing it — checking the geometry, slicing it in Bambu Studio, and iterating through failed prints until it holds together.
- Blender
- Cinema 4D
- 3ds Max
- FDM printing
- Bambu Studio
- Physical iteration
MY OPERATING SYSTEM
My operating system
Fast does not have to mean careless. I use AI agents aggressively, not blindly. I keep the work grounded in project context, real constraints, review, and deployment.
- 01
Find the real problem
I reduce a vague idea to one useful user action and one clear outcome.
- 02
Build the system, not only the screen
I map data, states, edge cases, AI behavior, backend boundaries, and the path to deployment.
- 03
Let agents execute, keep the judgment calls
Claude Code and Codex handle the mechanical parts — scaffolding, repetitive edits, boilerplate. Every change still gets reviewed against project memory and constraints before it ships.
- 04
Iterate in public reality
Real devices, real API limits, weird edge cases, store requirements, broken generations, and failed prints are part of the process.
- 05
Ship, observe, refine
Push, deploy, test, collect feedback, fix what matters, repeat.
AGENT-NATIVE, PRODUCT-LED
Agent-native, product-led
Every change follows the loop below: an intent, project memory, one agent turn, a check on a real device or browser, then Git and deploy. Feedback from that becomes the next intent.