Life sciences researchers can now access NVIDIA's GPU-accelerated computing stack through natural language commands via Anthropic's Claude Science platform.

Claude Science, announced this week as an AI workbench for scientific research, integrates with NVIDIA BioNeMo Agent Toolkit to let scientists run end-to-end workflows by conversing with AI agents. The toolkit packages NVIDIA-accelerated capabilities as callable skills that Claude Science can select and execute automatically.

The integration connects researchers to NVIDIA compute resources deployed anywhere, bringing accelerated models, libraries and NVIDIA NIM microservices directly into their research environment. Scientists can describe tasks like analyzing genomic sequences or predicting protein structures in plain English, with Claude Science orchestrating the computational work through specialized agents.

Accelerating Scientific Workflows

The toolkit addresses a key bottleneck in AI-driven research: computational speed. NVIDIA Parabricks accelerates genomic analysis from hours to minutes, while RAPIDS-singlecell compresses a 1.3-million-cell preprocessing workflow from 52 minutes to 25 seconds.

Cheminformatics operations see even larger gains through nvMolKit, which accelerates similarity searches and conformer generation by up to 3,000x. These speed improvements let AI agents integrate complex analyses into real-time reasoning loops rather than running them as separate batch jobs.

One example workflow involves generating cancer inhibitors. A scientist starts with a known cancer-causing antigen mutation and asks Claude to design potential inhibitors. The system then accelerates high-throughput inhibitor prediction, optimization and validation through integrated NVIDIA NIM microservices.

18 of the top 20 pharmaceutical companies already use NVIDIA BioNeMo for AI-enabled research across drug discovery, genomics, medical imaging and protein engineering, according to the company.

The BioNeMo Agent Toolkit includes open models for core biomolecular capabilities, all accelerated by NVIDIA libraries. These models are packaged as enterprise-ready NIM microservices with pre-integrated software stacks optimized for high-performance inference.

The toolkit is harness-agnostic, allowing the same scientific skills to work across different agent frameworks and research platforms. It's available now through NVIDIA developer resources and GitHub.

Claude Science enters public beta today, with Anthropic inviting researchers to provide feedback on additional domain specialists and integrations needed for their work.