AI4Simulation

At NXAI, we firmly believe that AI is on the cusp of disrupting industry-scale simulations. In order to spearhead this transformation, we undertake cutting-edge research with a committed team and extensive computational resources overseen by Johannes Brandstetter. Simply put, we want to change the way simulations are done in industry.

The Goal

Over recent years, especially with the advent of Large-Language Models, it has become evident that scalable paradigms are changing the world by storm. After language and compute vision, foundation models are starting to spread over to various disciplines, such as weather modeling or protein design. Having the in-house expertise on scaling-up deep learning models, NXAI aims to develop foundation for every-day engineering processes, and in doing so, disrupt how our industry works.

Simulations in Seconds

Every day, thousands upon thousands of computing hours are spent for modeling turbulences, simulations of fluid or airflow, heat transfer in materials, traffic flows, and much more. Many of these processes follow similar basic patterns but still require different and highly specialized software for simulation. More problematic is the fact that costly simulations for various parameter settings must be redone from scratch, making it capital and time-intensive. That is precisely what we're changing now.

Deep learning techniques are ready to develop models that perform simulations in seconds instead of days or even weeks. The hardware is prepared to process high-resolution inputs at an industrial scale, such as 3D meshes or images, thereby setting the stage for training deep learning models on a large scale. Many of the processes mentioned above have a common underlying dynamic – similar to how different languages share a common structure and grammar. Simulation data is abundant; we use the right kind, and plenty of it.

What Makes the Field of Simulation Research So Exciting?

AI-driven simulations are on the brink of transforming our industry and the operational methods of companies in general. Instead of heading to the lab, we'll work with digital twins, allowing us to examine physical or technical systems in real-time. But that's not all: AI-driven simulations will unlock the potential to discover innovative designs beyond our imagination, uncover hidden dynamics, and enable experimental cycles at vastly accelerated rates.

About Johannes Brandstetter:

As a figure in AI research, Johannes Brandstetter returned to Linz from Microsoft/Amsterdam in December 2023 to develop "the tools that will distinguish our industries" alongside his mentor Sepp Hochreiter at JKU Linz.

"We're starting with particle-based simulations – concepts crucial for a wide range of simulation tasks, from modeling multi-fluid systems and fluid-material interactions to calculating forces in silos and predicting the spread of chemical outbreaks. Deep learning allows us to scale to millions of particles, thus creating new tools for conducting real-time simulations of processes that traditionally took hours or many days. As we expand the scope of our modeled processes and generate more data, our models become increasingly complex. Eventually, these models will reach a point of generalization that surpasses human understanding of the underlying principles. This marks the beginning of an exciting phase that ushers in a new paradigm in simulation technology."