Started programming at 16
PHP first, then iOS. The through-line stayed the same: build the tool, ship the app, reduce the distance between idea and working system.
Experience
I build AI systems from the inside of real work.
iOS first. AI now. Ex-Uber. Amsterdam. Sixteen years shipping software, with the last few spent turning the AI wave into working products, tools, and team workflows.
Platform, server-driven UI, payments
Uber taught me the boring part of serious engineering: ownership, cross-team systems, build complexity, and reliability where a small percentage error can become a real incident.
Hackathons, LangChain, early agents
I went deep when the first practical AI tooling wave hit: hackathons, LangChain experiments, financial-analysis agents, early workflow automation. Most of it was messy. That was the point — the edge was moving faster than polished frameworks could explain it.
Stable Diffusion, ComfyUI, local media pipelines
I moved from image-generation curiosity into reproducible local pipelines: Stable Diffusion-family models, ComfyUI graphs, video workflows, model management, GPU failures, and the operational layer that turns a demo into something repeatable.
WhisperBoard
WhisperBoard is my proof that mobile AI is not only API wrappers. On-device transcription, privacy-first architecture, App Store delivery, 50k+ downloads, 4.8 rating, and 1,035 GitHub stars.
Codementor and practical debugging
Codementor keeps me close to real demand: people stuck on iOS, SwiftUI, AI integrations, architecture, and debugging. The valuable work is rarely picking a model. It is finding the constraint that makes the system fail.
Agent-driven development before it became the default
I use agents on real codebases daily. That forced me to care about repo legibility, context files, MCP tools, test loops, visual feedback, review gates, and architecture that agents can actually navigate. The prompt is not the system. The workflow is.
AI architect for teams that need production, not theater
Today I combine lead iOS work with internal AI enablement: helping engineering teams adopt AI safely, pick the right workflows, and avoid turning every prototype into an unmaintainable shrine to a demo that once worked on a laptop.