Webinar
AI for Science: From the Lab to the Frontier
Join NVIDIA and leading academic partners for a three-part webinar series spanning structural biology, climate science, and computational chemistry. Each session stands alone—attend one, two, or all three. The series pairs NVIDIA’s latest AI platforms with the scientists and engineers using them to showcase what’s possible today and where the field is headed next.
In this series, you’ll learn about:
- Structural Biology: See how AI is predicting protein structures and designing drug candidates with the NVIDIA BioNeMo™ Agent Toolkit.
- Climate Science: Discover how NVIDIA Earth-2 enables kilometer-scale climate simulation to help researchers tackle real-world climate risk.
- Materials Science: Explore how NVIDIA ALCHEMI™ is helping redefine atomic-scale simulation by accelerating the next generation of machine-learning interatomic potentials.
In partnership with the European Commission, hear from Liviu Știrbăț on their AI in Science initiative — empowering researchers with AI for scientific discovery.
Agenda
-
September 23
-
September 30
-
October 6
9–10 a.m. PT
10–11 a.m. CEST
AI for Science: AI-Driven Drug Discovery With BioNeMo
Discover how the BioNemo Agent Toolkit equips researchers and engineers with production-ready AI skills for two of the most impactful challenges in structural biology. In this session, we'll walk through how to deploy state-of-the-art models for protein structure prediction and demonstrate how generative AI can design high-affinity protein binders, improving the quality of drug candidates. Whether you're new to AI-driven drug discovery or scaling an existing pipeline, this session bridges cutting-edge research with practical application.
9–10 a.m. PT
1–2 p.m. CEST
AI for Science: Building With NVIDIA Earth-2: From Open AI Models to Real-World Climate Applications
Climate and environmental risks are intensifying—from extreme weather events disrupting infrastructure to wildfire smoke blanketing entire regions. Decision-makers in government, energy, insurance, and public health need high-resolution, localized forecasts to act, but traditional numerical simulation is too slow and too expensive to run at the resolutions that matter. NVIDIA Earth-2 provides a family of open AI models and tools that researchers can train on their own data and deploy for their specific domain. In this session, Niall Robinson introduces the Earth-2 stack, and Professor David Topping from the University of Manchester shows how his group used it to build street-level air quality forecasts, compressing days of compute into seconds.
9–10 a.m. PT
10–11 a.m. CEST
AI for Science: Accelerating Materials Discovery With Unconstrained Machine Learning Potentials
For decades, simulating how materials behave at the atomic scale meant choosing between accuracy and speed. Machine learning is dissolving that tradeoff, but a deeper question has emerged: Do AI models for materials need to be built around the laws of physics, or can they learn those laws directly from data? This session explores that question through the lens of Point Edge Transformer (PET), an unconstrained architecture that learns physical symmetries from data rather than enforcing them by design, and PET-MAD, its universal instantiation trained on the MAD (Massive Atomistic Diversity) dataset to simulate virtually any material across the periodic table, from solid-state battery electrolytes to organic molecules. The session also highlights how recent collaborations with NVIDIA ALCHEMI enabled PET acceleration through GPU-accelerated, batched kernels and floating-point emulation of SGEMMs, and how the broader scientific community can leverage these tools to accelerate materials discovery.
Speakers
Join Us at GTC Berlin
October 20–22
Connect with fellow developers, researchers, and industry leaders on AI infrastructure. Sessions will span the entire five-layer AI stack, from energy, chips, and infrastructure to open models and physical AI applications that empower every industry and organization to build AI on their own terms.
Academic, Government, and Non-Profit Discounts Save 25% on Conference passes when you register with a university, government, or nonprofit email address.
