Michael Tarekegn

I'm currently doing ML research in the math department at UCLA, where I study linear approximations to self-attention, and at Cal State LA, where I build spatiotemporal models for a NASA-funded air quality project. I'm also cofounder and CTO of the AI & Machine Learning Initiative, which teaches AI to teens.

My research interests sit where math, physics, and machine learning meet: why models train the way they do, and where cheaper approximations hold up or break down.

This year I had work on optimizer and precision interactions in PINNs accepted to the NeurIPS OPT workshop, and a paper on multilingual LLM attribution named a best paper finalist at the IEEE BigData High School Symposium. You can find the rest in more about me or on my GitHub.