I build software, end to end.

I'm early in my freelance work, and I'd rather be upfront about that than pad a page with claims.

What I do have is code you can read. Every project below is public: the commits, the mistakes, the refactors. That's a more honest signal than a logo wall, and it means you can judge how I actually work before spending anything.

I build because I want to know how things work underneath. Quietude began as a question about whether a voice assistant could work without sending anything to a server, and answering it meant learning speech recognition, speech synthesis, face recognition and how to keep a model that crashes from taking the whole application down with it. Camaloop began as a question about where a webcam feed actually lives on Linux. Most of what I know came from picking a problem slightly beyond me and refusing to put it down.

The machine learning side holds my attention most right now. Quietude runs two speech models at once. A small streaming one always listening for a wake phrase, and a quantized Whisper that transcribes properly once it's awake. It speaks through ONNX voice models, and recognises its user with a face classifier trained on their own machine. It can call a hosted model if you give it a key, but the interesting part is how much of it works without one. I want to keep building in that direction.

I'm in the first year of an IT degree at Curtin Mauritius, so there's a lot I haven't met yet, and I'd rather say so now than discover it halfway through your project. I work across the stack: web applications, automation, internal tools, and the deployment and maintenance side that usually gets skipped. On larger builds I work alongside another developer, so there's capacity for more than one person's worth of project. Being new means I'm responsive, priced accessibly, and more interested in the problem than the invoice. It doesn't mean I'll take on something I can't deliver: if a project isn't a fit, I'll say so.