NVIDIA is introducing three new software tools at ISC Hamburg to accelerate AI-driven scientific research. cuPhoton is a reference code for processing petabytes of astronomical and photon data, achieving up to 14,900x faster FITS image loading for the Rubin Observatory's LSST survey. DAQIRI is a high-performance networking library for real-time data streaming from fast detectors, used at CERN's ATLAS experiment to analyze collision data that would otherwise be discarded. ALCHEMI is a suite of NIM microservices for chemical and materials discovery, enabling GPU-accelerated molecular dynamics and geometry relaxation simulations. Lila Sciences demonstrated 50x faster materials screening and 6x training/inference speedups using ALCHEMI. All tools are part of the NVIDIA CUDA-X ecosystem and are available or coming soon via GitHub, PyPI, and the NGC catalog.
