I teach machines to see.
Machine Learning + Full-Stack / Dubai, UAE
Software is easier to make than ever. Libraries are everywhere, models are increasingly accessible and AI can help turn an idea into working code before you've finished your coffee.
What interests me is what comes before and after that moment: choosing the right problem, understanding why a system behaves the way it does and making something that feels considered.
That's part of what drew me towards machine learning. It isn't only about writing the code that makes a model run. A lot of the work happens in the questions around it: deciding what to measure, understanding the data, interpreting failure, figuring out why a prediction happened in the first place and deciding what that prediction should actually mean.
And somewhere along the way, I started caring about another kind of decision too — how an idea is presented once it leaves the codebase. I like design, and a portfolio is the one place nobody can overrule me on it.
Two people can be given the same tools and make completely different things. There's always more than one way to interpret a problem, so as independent as I like to be, sometimes the best thing you can do is listen to someone else's ideas. Or steal. Ethically, of course. Like how people get inspired by historical buildings and recreate them to fit modern needs. Ethical stealing.
I don't pick projects by job title; I pick them by what I want to be able to make. That's why my work runs from computer vision to LLM red-teaming to a UX study. Some of it turns out well, some gets left forgotten, but all of it teaches me something. Exploring the tech world is far from linear in the end.
And if I've caught your interest after all that
Recent projects
View all
A chess trainer with a C++ engine compiled to WebAssembly, so the whole thing runs in your browser.
A team project that cross-checks a claim's damage photos, the claimant's story and the police report.
An interactive essay that pits a one-line rule against machine learning on 395 real students.
Finds the document in a messy photo, straightens and cleans it, then works out what it's looking at.
A chess trainer with a C++ engine compiled to WebAssembly, so the whole thing runs in your browser.
A team project that cross-checks a claim's damage photos, the claimant's story and the police report.
An interactive essay that pits a one-line rule against machine learning on 395 real students.
Finds the document in a messy photo, straightens and cleans it, then works out what it's looking at.