I'm Charl Roux — 7 years shipping enterprise frontend systems, deepest in Angular but not defined by it: React on client delivery since 2019, recently building browser-native AI with WebLLM/WebGPU, and shipping React on top of a FastAPI backend I wrote myself. I design reactive state, lead complex migrations, and ship accessible production UIs that hold up at scale.
Five projects across reactive streams, finite-state UIs, live device control, on-device AI, and large-scale virtualization — four run entirely in the browser; one talks to a real FastAPI backend over REST and WebSocket.
Operator console for a virtual BACnet air-handling unit — React front to FastAPI back, end to end. React hooks drive a shared polling + WebSocket data layer against a backend I wrote: live point overrides, priority arrays, and alarm acks against a real device model instead of mock data.
Real-time Binance feed streamed into a 10,000-row virtual grid. An adjustable
RxJS bufferTime window decouples render pressure from the socket —
500+ events/sec batched into one update per interval; an in-browser LLM
auto-analyzes any move over 2%. No server, no backend, no API key.
Invoice lifecycle app where draft → pending → paid is a finite state machine at the service layer — invalid transitions are structurally impossible.
Generates a strong password and maps each character to a memorable word. Runs entirely in a Web Worker via a local LLM — no server, nothing leaves the device.
1,000+ listings virtualized on a single page. Hybrid vector + BM25 search via Orama and all-MiniLM-L6-v2, running on WebGPU; filter state synced to the URL.
A pragmatic, opinionated stack — picked because it scales for teams, not because it's fashionable.
I'd love to hear about your project and how I could help. Drop a note via the form, or reach me directly — I read everything and reply within a day or two.
