Welcome to the homepage of NVIDIA’s Fundamental Generative AI Research (GenAIR) group, led by Arash Vahdat. We are part of the Learning and Perception Research (LPR) organization and mostly focus on generative learning and its applications in different areas. We are particularly interested in:
We are currently looking for outstanding candidates to join our team as:
For general inquiries, you may email your resume to genair-openings@nvidia.com. Kindly note that we may not be able to respond to all inquiries via email.
Jan 2026 - We released NVIDIA FastGen a plug-and-play library that distills slow diffusion models into 1-4 step generators. It currently supports 12 models and 13 state-of-the-art distillation methods. Learn more at this blog.
Oct 2025 - Our biomolecular design models including GenMol, ReaSyn, and La-Proteina are now offered as part of NVIDIA Clara Open Models, following the announcement at GTC DC. Read the blog.
Sept 2025 - ReaSyn, a generative model of molecular synthetic pathways, has been released. Check out the paper, blog, or code. ReaSyn got an Oral presentation at ICLR 2026!
Feb 2025 - CorrDiff, a diffusion model for weather super-resolution, has been accepted to Nature Communications Earth & Environment and is now available in NVIDIA Earth-2. Check out the paper, blog, or online demo.
Feb 2025 - We’ll present 11 papers at ICLR 2025 including two Oral presentations. Can’t wait to see you in Singapore!
Jan 2025 - GenMol a generalist generative model for molecular tasks was announced at J.P. Morgan conference and is now available in NVIDIA BioNemo. Check out our paper, blog, or online demo for GenMol.
Dec 2024 - We’ll present 5 papers at NeurIPS in Vancouver!
Dec 2024 - We’ll present 2 papers at SIGGRAPH Asia in Tokyo!
Nov 2024 - The GenAIR website goes live!!