Ahmed Ali Abdellatif
Full-stack developer · Cairo, Egypt
I got into software because of a website. Something animated, doing things I couldn't figure out just by looking at it, and I wanted to know how.
How I got here
That wasn't what I was studying. I studied computer science at MSA University, and for most of the degree I wanted to go into games. Real-time graphics, systems that react to you. Egypt's game industry is small enough that building a career on it felt like a gamble. I learned enough Unity to understand what I was giving up, then went after the math that sits underneath the graphics instead.
Image processing, then pattern recognition, then neural networks. That last one is where it stopped feeling like coursework. My graduation project came out of it: a model that predicts brain activity from visual stimuli. You give it an image and it estimates how a brain would respond to seeing that scene. I built it as a twelve-layer encoder-decoder transformer, and it worked well enough that I expected to spend my career in machine learning.
Why I build interfaces
In my third year I went back to the thing that started it. People were building interfaces I didn't think a browser could do, with motion and depth and type doing actual work, and the reason I'd been pulled toward graphics turned out to be the same reason I'd been pulled toward neural networks. Both of them look impossible until you can see the mechanism. I like being on the inside of that.
So I taught myself the stack my degree never covered, on my own time, while I was still finishing the degree. That's roughly how I still work. React and Next.js at the front, C#/.NET and PostgreSQL behind it, and I host most of what I build myself instead of handing it to a platform. Knowing what happens between a request and a rendered frame is part of the design work. It's what keeps a design intact once real traffic hits it.
How I work
I've liked puzzles since I was a kid. Chess, riddles, anything where you have to find the shape of the thing instead of following a method you can look up. That's most of what I bring to a problem. I'm competitive, though mostly with myself: the thing I'm trying to beat is usually whatever I built last. It makes me slow to call something done, and difficult to argue out of a detail that's nearly right.
What I'm aiming for is that whatever I build makes someone look at it twice instead of once.