Simulation for accelerated discovery

Our Accelerated Discovery Technology Foundations work has four major technology stacks: Knowledge, Simulation & Modelling, Generative, and Lab of the Future (also known as RXN). Together, they aim to enhance the scientific method.

The four major functional areas for driving a better, more impactful, scientific method. (Credit A. Annunziata)

These four areas each have a toolkit that we have built:

1. DS4SD – A DeepSearch toolkit
2. ST4SD – A Simulation toolkit
3. GT4SD – A Generative model toolkit
4. RXN – A Tool for digital chemistry in the cloud

The major focus of the Simulation Team under the AD strategy is to build, validate and extend ST4SD with state of the art methods in quantum, hybrid cloud and AI – including Foundation Models.

ST4SD

Modelling and simulation afford scientists the ability to test hypotheses in silico, which allows for enhanced scalability and throughput at a lower cost. Simulation workflows are however notoriously complex, with complexity occurring both within workflow design and deployment. There are a multitude of datapoints, methods, and systems to choose from. Choose wrong at any point, and your entire experiment could be flawed before it's begun, costing dollars and hours of computing. Knowing which methods and systems to deploy is necessary, though not sufficient for deriving value – you must also be able to execute the required code bases reproducibly, often on HPC systems – which are known to be complex environments.

The Simulation Toolkit for Scientific Discovery (ST4SD) allows development and execution of virtual-experiments, querying of their output and sharing them with the community along with AI acceleration capabilities for virtual-experiment campaigns. The primary execution environment for ST4SD is OpenShift/Kubernetes, although executing in a classic HPC environment like LSF is also possible.

ST4SD operates with the mission to transform the experience of high-end technical computing so that those that couldn’t can, and those that could, can do more.To enable this, we looked to develop a state-of-the-art automation and workflow technology, and couple this with leading-edge science, enhanced by powerful AI.

Modelling and simulation remains a key technology for scientific discovery. In addition to embracing new paradigms for calculation, the team is actively pursuing the use of AI to accelerate modelling and simulation workflows. Our accelerated simulation technology team have three specific focus areas:

1. Simulation tech for science. Using the latest technology and methodologies to develop and enable cutting edge scientific simulations.

2. AI enriched simulation. Driving differentiation using AI to accelerate time to insight by both accelerating individual simulations using data-driven surrogate modelling, and the intelligent orchestration of simulation campaigns through Bayesian optimization.

3. Simulation workflow technology. Developing differentiated workflow technology to integrate the latest simulations with the latest AI enrichment in a hybrid cloud infrastructure.