As part of our ongoing efforts to foster dialogue on cutting-edge artificial intelligence research, the ELLIS Madrid Unit is pleased to co-organize an invited talk by Dr. Morteza Mardani, Principal Scientist at NVIDIA Research, on Monday, January 19, 2026.
The talk, entitled “Steering Diffusion Models for Generative AI”, will take place from 11:00 to 12:00 at the Salón de Actos, Edificio de Gestión, on the Fuenlabrada Campus of Universidad Rey Juan Carlos (URJC). For those unable to attend in person, the event will also be accessible online via Microsoft Teams.
Diffusion models have rapidly become a cornerstone of modern generative AI, driving progress across vision, scientific computing, and natural language processing. In this talk, Dr. Mardani will discuss how large-scale diffusion and foundation models can be steered at test time to address complex downstream tasks. He will present practical approaches based on guidance mechanisms and reinforcement learning, highlighting their benefits, limitations, and real-world applications.
Dr. Mardani is a leading researcher in generative and statistical learning. At NVIDIA Research, he leads work on generative AI algorithms, while also serving as a visiting researcher at Stanford University and a Distinguished Industry Speaker for the IEEE Signal Processing Society. His academic background includes a PhD in Electrical Engineering from the University of Minnesota, as well as research appointments at Stanford University and UC Berkeley. His contributions have been recognized with several awards, including the IEEE Signal Processing Society Young Author Best Paper Award.
This invited talk is supported by the IEEE Signal Processing Society Distinguished Speaker Program (IEEE Spain Section), with ELLIS Madrid Unit and Universidad Rey Juan Carlos as secondary sponsors.
For additional information or specific requests, please contact gr_inv.dssp@urjc.es.
We look forward to welcoming researchers, students, and practitioners interested in generative AI and diffusion models to this event.