Let’s connect

I’m always happy to talk about interpretability, tensor methods, and safe AI — collaborations, questions about the papers, or a coffee if you’re at Mila.

farzaneh.heidari@mila.quebec