At the ISC conference in Hamburg this week, NVIDIA introduced new software designed to accelerate AI for science, spanning chemistry, materials discovery, and the search for dark matter. The core offerings are the NVIDIA DAQIRI library, new NVIDIA ALCHEMI NIM microservices, and the NVIDIA cuPhoton reference code, which is coming soon. These tools are part of NVIDIA CUDA-X, a collection of tools and libraries that deliver significantly higher performance across application domains including AI and high-performance computing. The performance gains are substantial and have real practical impact, enabling scientists across disciplines to generate data and insights with instruments and surveys faster than ever before.

The cuPhoton reference code accelerates the loading, reading, processing and analysis of FITS data, the standard astronomical file format used by observatories and telescopes. When running on NVIDIA GB200 NVL72 systems, cuPhoton demonstrated extraordinary performance improvements in early access testing with the Rubin Observatory's Legacy Survey of Space and Time. It accelerated loading and reading of FITS images by 14,900x and enabled up to 8,400x faster signal processing and analysis using 32 NVIDIA Grace Blackwell superchips. This acceleration directly benefits the LSST camera, described as the largest digital camera ever built, which captures images of billions of distant galaxies as well as closer, faint objects that do not reflect much light. Princeton University collaborated with NVIDIA to develop cuPhoton, and both Princeton and Harvard University plan to use it for processing and analysis of massive data collected from observatories and dark energy surveys.

NVIDIA DAQIRI, short for Data Acquisition for Integrated Real-time Instruments, is a high-performance networking library that streams data from fast detectors and sensors into NVIDIA software. Traditional systems are constrained by fixed hardware and can lose data when instruments produce it faster than they can save it. DAQIRI solves this by handling the data stream as it arrives. The A-GHOST research project, developed by scientists from CERN, the University of Chicago, and University College London within the CERN openlab framework, demonstrates this capability by using DAQIRI to run AI in real time on collision data from the ATLAS Experiment at CERN. A-GHOST analyzes data that would normally be rejected by ATLAS due to storage constraints, over 99 percent of the raw collision data, allowing it to detect potentially interesting signals that would otherwise be lost.

NVIDIA ALCHEMI is a collection of domain-specific microservices and toolkit for accelerating chemical and materials discovery, with applications spanning battery materials, catalysts, OLED displays, beauty products, and more. In March, NVIDIA released two ALCHEMI NIM microservices for batched geometry relaxation and batched molecular dynamics, AI-accelerated tools that allow researchers to simulate millions of molecules and materials simultaneously. The geometry relaxation tool identifies their most stable structures, while the molecular dynamics tool simulates how they move over time. ALCHEMI is expected to soon include a microservice for the Vienna Ab initio Simulation Package, a widely used tool that enables researchers to run materials simulations with higher GPU throughput. By running multiple VASP calculations on a single GPU with NVIDIA Multi-Process Service, the microservice achieves a 3x speedup for geometry optimization, the process of finding the most stable arrangement of atoms in a material.

Lila Sciences, a company building a scientific superintelligence platform and autonomous lab for life sciences, chemistry, and materials science, collaborated with NVIDIA on a high-fidelity magnet simulation using ALCHEMI, which was demonstrated at NVIDIA GTC San Jose in March. Lila Sciences accelerated high-throughput materials screening by 50x using the ALCHEMI NIM microservice for batched geometry relaxation, identifying stable candidates with higher chances of successful synthesis. The company achieved a 30 percent speedup in calculating magnetic properties for shortlisted candidates using the ALCHEMI VASP microservice in early access. The performance improvements compound further: ALCHEMI's specialized kernels for TensorNet delivered Lila Sciences a 6x speedup in training and inference while reducing memory usage by 3x, enabling simulations that previously required weeks to complete in just days. Rather than running one experiment at a time, this approach evaluates multiple materials simultaneously in GPU memory and can be generalized across diverse use cases.

Lila Sciences also accelerates scientific discovery using the broader NVIDIA stack, including NVIDIA Megatron-LM and NVIDIA Nemotron for training, the Nemotron 3 Nano and Nemotron 3 Super open models, NeMo RL and NeMo Gym libraries, NVIDIA BioNeMo for molecular generation, NVIDIA Triton and NIM microservices for inference serving, and NVIDIA Omniverse libraries for digital twins. Andy Beam, cofounder and chief technology officer of Lila Sciences, stated: "The work showcases using a powerful computing stack assembled to accelerate discovery at a scale no individual scientist could achieve alone."

The NVIDIA ALCHEMI Toolkit and Toolkit-Ops are available for download from Github and PyPI, while ALCHEMI NIM microservices are available from the NVIDIA NGC catalog. The ALCHEMI NIM microservice for VASP is expected to be available later this summer. DAQIRI is now available on Github. cuPhoton is expected to be available this summer.