Nima Fathi is a visiting researcher at ServiceNow on the multimodal foundation models team. He built one of the first diffusion-based language models and proposed a framework to unify large language models and diffusion-based language models for efficient generation.
As a research scientist at Mila – Quebec AI Institute, he designs and implements multimodal diffusion architectures spanning discrete language, image, and video domains. He also leads the creation of agentic AI solutions in healthcare and drives open-science impact by contributing code and preprints on conditioning strategies, latent-space manipulation techniques, and safety-first agent policies.
