Mélissa Colin

I'm an engineering student at ENSEIRB-MATMECA, on the engineer-doctor track. What drives me is rethinking model architectures: making them more inspired by living systems, able to build their own "model of the world" and more efficient than Transformers. Right now, I'm a Mitacs Globalink research intern at the University of Alberta's Vision & Learning Lab, working on physics-grounded 3D motion generation. And my longer-term goal is to become a Research Scientist at Google.

From a village in Gironde to AI research

I grew up in a rural village in Gironde, in a working-class family. As a kid, I was told I wasn't made for studying, but for working. I taught myself to code: Scratch at 12, Python at 13, then a technical high school. What captivated me, very quickly, was algorithms and the way a machine can learn, far more than websites or games.

The spark came during an internship at Cali Intelligences, a computer-vision startup. I threw myself into it with a conviction that never left me: rather than stacking more layers, I wanted to contribute to the very foundations of AI by rethinking model architectures. I moved up one step at a time, ranking first at each stage, until I was admitted to ENSEIRB-MATMECA on merit and ranked 3rd out of 93, then earned a Mitacs Globalink scholarship to do research at the University of Alberta.

Why I do this

At heart, my drive isn't only scientific. I want to prove it's possible: so that another girl from a modest background can look at me and think, "if she could, then so can I." Science is the tool, opening the way is the goal. And I hold one red line: never to put my work in the service of surveillance or military uses. My motto is a line from Stephen Hawking: "Intelligence is the ability to adapt to change."