AQChemSim

AQChemSim is a chemicals and materials discovery platform developed by SandboxAQ that leverages Large Quantitative Models (LQMs) derived from AI and physics-based approaches to revolutionize the way chemicals and materials are designed and optimized. Built with NVIDIA DGX™ Cloud, the platform is designed for researchers, R&D teams, and enterprises across industries including energy storage, specialty chemicals, and advanced manufacturing who need to accelerate innovation beyond what traditional computational methods allow.

AQChemSim combines data-driven insights with high-accuracy simulations to help customers discover and optimize materials against desired properties including reactivity, synthesizability, kinetics, toxicity, mechanical performance, and electronic characteristics. Its core differentiated capabilities span LLM-based chemical data and information extraction, cloud parallelization, automated high-throughput materials discovery workflows, and guided multi-objective generative chemistry — all coupled with LQMs that leverage Density Functional Theory (DFT), Iterative Full Configuration Interaction (iFCI), Generative AI, Bayesian Optimization, and Chemical Foundation Models.

Core Application Areas

• Energy Storage: From Li-ion batteries to next-generation solid-state batteries, AQChemSim enables discovery and optimization of materials that enhance battery life, improve capacity and efficiency, and support broader decarbonization efforts.
• Chemicals and Catalysts: The platform designs and refines advanced catalysts by simulating reaction mechanisms and optimizing reactivity and selectivity of chemical processes, supporting the development of sustainable compounds and formulations.
• Alloys: By leveraging large material datasets and high-accuracy simulations, AQChemSim designs and optimizes advanced alloy compositions for applications requiring durability, lightness, and sustainable processing.

Key Use Cases and Demonstrated Breakthroughs

• Battery Life Prediction (with NOVONIX): SandboxAQ's LQMs were trained on five years of ultra-high precision coulometry data to predict lithium-ion battery end-of-life (EOL). The approach achieved predictions in days rather than months, delivering 35x greater accuracy and requiring 50x less data than traditional AI models, reducing overall testing time by 95% and lowering R&D costs.
• PFAS Chemistry Simulation (with AWS, Intel, and Accenture): SandboxAQ turned the AWS cloud into a massively distributed supercomputer, enabling near-exact simulation of bond breaking in toxic PFAS molecules using over one million vCPUs — a first in the history of chemistry. This breakthrough supports PFAS remediation, sustainable PFAS replacement design, green chemistry, battery materials, and sustainable agriculture.
• Next-Generation Catalyst Design (with DIC and AWS): Using QEMIST Cloud, SandboxAQ simulated the largest organometallic catalyst computed at near-exact accuracy to date, breaking the traditional computational limitation that restricted high-accuracy quantum chemistry to small model systems. This enables AI-driven materials design and drug discovery at industrially relevant molecular scales.
• NVIDIA-Accelerated Quantum Chemistry: In collaboration with NVIDIA, SandboxAQ combined its LQMs with the CUDA-accelerated Density Matrix Renormalization Group (DMRG) algorithm, achieving computing speeds more than 80x faster than traditional 128-core CPUs. This enabled unprecedented calculations for complex biochemical systems including transition metal metalloenzymes, with implications for medicine, energy, and consumer products.

Scientific Advisory — The SandboxAQ Braintrust

• Dr. Dane Morgan, Professor of Materials Science and Engineering, University of Wisconsin
• Dr. Eva Zurek, Professor of Chemistry, University of Buffalo
• Dr. Chris Marianetti, Professor of Materials Science, Applied Physics, and Applied Mathematics, Columbia University
• Dr. Nikhil Gupta, Professor of Mechanical Engineering, New York University
• Dr. Paul Zimmerman, Professor of Chemistry, University of Michigan
• Fei-Fei Li, Sequoia Professor of Computer Science at Stanford University, former VP/Chief Scientist of AI/ML at Google Cloud
• Dr. Adrian Roitberg, Professor of Chemistry, University of Florida

AQChemSim is deployed as a cloud-native platform, with demonstrated integrations with AWS, NVIDIA DGX™ Cloud, and Intel infrastructure. The platform has been applied in partnership with commercial organizations including NOVONIX and DIC, as well as government programs such as a collaboration with the U.S. Army to support AI-enabled materials discovery for armored vehicle alloys and battery design. Product-specific offerings include AQCat for catalyst discovery and AQVolt for energy storage applications.