NVIDIA adds tools for faster scientific AI workflows

NVIDIA introduced new software at the ISC conference in Hamburg to accelerate AI-driven scientific computing across chemistry, materials discovery, astronomy and dark matter research. The CUDA-X additions include the DAQIRI data acquisition library, ALCHEMI NIM microservices and the forthcoming cuPhoton reference code, designed to shift workloads that once took hours or days on CPUs into real-time GPU pipelines.

CuPhoton targets multidimensional data from telescopes, X-rays and laser experiments. In early access, cuPhoton accelerated loading and reading of FITS images from the Rubin Observatory’s Legacy Survey of Space and Time by 14,900x and enabled up to 8,400x faster signal processing and analysis using 32 NVIDIA Grace Blackwell superchips. Princeton University collaborated with NVIDIA on cuPhoton, with Princeton and Harvard University planning to use it for observatory and dark energy survey data.

DAQIRI streams data from fast detectors and sensors into NVIDIA software for real-time processing. The A-GHOST project from CERN, the University of Chicago and University College London uses DAQIRI to run AI on ATLAS Experiment collision data that would normally be rejected, over 99% of it, because of storage limits.

ALCHEMI focuses on chemical and materials discovery through microservices and tooling for atomistic simulation workflows. Lila Sciences used ALCHEMI to accelerate high-throughput materials screening by 50x, improve magnetic property calculations by 30%, and gain a 6x speedup in TensorNet training and inference while reducing memory use by 3x. ALCHEMI tools are available through GitHub, PyPI and NVIDIA NGC, while the VASP microservice and cuPhoton are expected this summer.