SandboxAQ reports that its AQCat model now allows researchers to screen catalyst candidates at up to 20,000 times the speed of conventional physics-based methods, while maintaining comparable accuracy. Trained on 13.5 million high-fidelity quantum chemistry calculations spanning 47,000 catalyst systems, AQCat addresses a critical bottleneck in materials innovation; traditional laboratory methods typically evaluate fewer than 100 candidates per week.
“Catalysis touches nearly everything the global economy produces, yet the tools to discover better catalysts have barely changed in decades,” said Jack Hidary, CEO of SandboxAQ. The company is now making AQCat generally available in AWS Marketplace, extending access to its Large Quantitative Models for a broader range of industrial users.
AQCat: Large Quantitative Model for Catalyst Discovery on SageMaker
AQCat accounts for spin polarization, a physical effect critical for accurately modeling industrial metals like iron, nickel, and cobalt, expanding the scope of viable catalyst candidates beyond those accessible to older methods. This capability creates possibilities for advancements in green hydrogen production, sustainable aviation fuel, fertilizer manufacturing, and plastics recycling, addressing key challenges in industrial sustainability. The model was trained using 13. SandboxAQ has made AQCat generally available in AWS Marketplace, allowing enterprise research teams to deploy the model within their existing AWS infrastructure without requiring specialized computational scientists or custom coding.
This shift from bespoke scientific engagements to scalable commercial distribution enables users to compress research and development timelines and reduce risk across the entire innovation pipeline, the company says. “Highly efficient machine learning interatomic potentials such as AQCat will rapidly accelerate the evaluation of promising new materials and deepen our understanding of their complex transformations,” said Julia Yang, Assistant Professor, School of Chemical and Biomolecular Engineering, Georgia Institute of Technology. Conventional laboratory methods typically assess fewer than 100 catalyst candidates each week, creating a significant bottleneck in materials discovery; AQCat removes this limitation by enabling rapid computational screening of vast material libraries.
“By making AQCat available in AWS Marketplace, we’re putting a capability that once required specialized teams and supercomputers into the hands of any enterprise research team, on infrastructure they already trust. This is how we scale our Large Quantitative Models from breakthrough science into everyday commercial impact.” AQCat also supports six additional industrially relevant elements, barium, cerium, fluorine, lithium, lanthanum, and magnesium, not included in other datasets, further broadening its applicability.
Highly efficient machine learning interatomic potentials such as AQCat will rapidly accelerate the evaluation of promising new materials and deepen our understanding of their complex transformations.
Julia Yang, Assistant Professor, School of Chemical and Biomolecular Engineering, Georgia Institute of Technology
See today’s quantum computing news on Quantum Zeitgeist for the latest breakthroughs in qubits, hardware, algorithms, and industry deals.
