NVIDIA today announced the NVIDIA BioNeMo Agent Toolkit, which provides domain-specific tools and skills for the agenttic life sciences era.
Image credit: NVIDIA
This toolkit includes over a decade’s worth of NVIDIA life science libraries, tools, and open models that enable AI agents, scientists, and labs to work together by gathering evidence, reasoning about findings, running computational experiments, and recommending next-best steps to accelerate discovery.
It gives any agent or AI platform, from general-purpose assistants to specialized scientific agents, software platforms, and in-house biopharmaceutical systems, the tools they need to synthesize and summarize scientific knowledge, invoke models, evaluate results, make inferences, and take next actions.
This toolkit includes NVIDIA BioNeMo™, which leverages NVIDIA NIM™ microservices, NVIDIA Parabricks®, NVIDIA NeMo™, and NVIDIA Nemotron™ technologies, accelerated computing and skills to provide an open and trusted foundation for agenttic life sciences.
More than 50 leading companies are already using it to advance scientific discovery, leveraging agent-callable skills for tasks such as protein structure prediction, molecular docking, generative chemistry, genomic analysis, protein design, and biomarker discovery.
The frontier model is the brain. BioNeMo is a scientific toolbox. Together, they give AI agents the skills of a PhD research assistant and the speed of a supercomputer. For the first time, researchers can now build AI agents that understand scientific knowledge, use scientific tools, and perform scientific workflows. This is a new way of doing science that can dramatically accelerate discoveries across biology, chemistry, genomics, and medicine. ”
Jensen Huang, Founder and CEO of NVIDIA
Open modeling and research organizations such as the Arc Institute, the Open Molecular Software Foundation, and the Institute for Protein Design (IPD) at the University of Washington are collaborating with NVIDIA to use BioNeMo to advance frontier models and make them more accessible through agent-enabled workflows. Our collaboration with IPD has reduced the run time of state-of-the-art biodesign models, such as RosettaFold3, delivering 2x faster performance than previous generation models. Additionally, many additional applications are currently underway to accelerate protein design efforts, providing researchers with tools at scale and cost not previously possible.
“All the tools we have built for protein design are only powerful if scientists can access them efficiently.” said David Baker, professor of biochemistry at the University of Washington School of Medicine and director of the Protein Design Institute. “The next breakthrough in science will not come from a single discovery. It will come from iterative design and the speed of agents that can iteratively reason about the complexities of biology at a speed that is impossible for humans.”
Agent-enabled tools and skills for life sciences
Life sciences are one of the world’s most important scientific frontiers, with global scientific research and development reaching $3.8 trillion and annual pharmaceutical budgets approaching $300 billion.
Agenttic workflows help industries iterate faster while reducing costs and maximizing the probability of success. Toolkits that allow developers to transform general-purpose agents into life science agents in minutes allow researchers to run experiments faster, continuously learn from results, and close the loop between hypothesis and discovery, and some companies are extending this iteration to the physical laboratory.
Generic agents can be difficult to navigate scientific workflows efficiently, as they must infer the correct tools, inputs, outputs, and biological meanings along the way. The BioNeMo Agent Toolkit enables agents to invoke the right tools, more accurately interpret results, and gain scientific insights faster and more reliably.
NVIDIA is optimizing the entire BioNeMo platform by turning libraries, models, and frameworks into agent-callable tools.
This includes leveraging NVIDIA Agent Toolkit technologies such as NVIDIA Nemotron open models for inference foundations, the NVIDIA NeMo RL library for reinforcement learning, and NVIDIA NemoClaw™ blueprints for secure private agents that can reason across tasks, invoke tools, and continuously interact with data.
NVIDIA NIM microservices help agents invoke models to perform tasks. The NVIDIA OpenShell™ runtime provides a controlled executable environment.
Toolkit components enable agents to complete workflows such as:
- Virtual screening: Agents help researchers identify small molecule drug candidates by generating and screening compounds, docking them to targets, predicting binding strengths, and filtering for drug-like properties. Agents can then output which candidates should be prioritized, reducing screening timelines from days to minutes.
- Genomic analysis and target discovery: Agents help researchers transform raw sequence data into prioritized genetic insights and biological targets. NVIDIA Parabricks accelerates alignment and variant calling, genome-based models score the impact of variants, and agents rank the most disease-relevant candidates for further study.
- Protein binder design: Agents help researchers computationally design and validate candidates before they begin work, compressing traditional labor-intensive design efforts.
- Deep biomedical research: Agents connect real-world data to inference models to improve efficiency and accuracy in a variety of scientific and clinical development processes, including literature review, protocol creation, clinical trial screening, and pharmacovigilance using the NVIDIA Biomedical AI-Q Research Agent.
- Medical image analysis: Agents help researchers process, segment, synthesize, and reason with medical image data to support biomarker discovery and accelerate evidence generation across research workflows.
Building a life sciences ecosystem with NVIDIA BioNeMo
Companies across the technology and life sciences ecosystem are using this toolkit to power agent workflows.
Frontier labs and scientific agent builders such as Anthropic, Edison Scientific, Lila Sciences, OpenAI, and Owkin integrate with BioNeMo to help agents transition from answering questions to completing scientific work. NVIDIA’s fast models and analytics libraries accelerate your time from hypothesis to insight.
Scientific data and workflow platforms from Benchling, Certara, Databricks, Snowflake, and Seqera use the BioNeMo Agent Toolkit to connect data systems with AI-powered science. BioNeMo’s skills help agents query biological and chemical datasets, prepare model-ready inputs, initiate reproducible workflows, analyze outputs, and return insights directly within a platform that scientists and data teams already use routinely.
Diagnostic and pharmaceutical companies such as Lilly and Natera use the BioNeMo Agent Toolkit to extend repeatable drug workflows across discovery, translational research, and clinical insights.
AI-native biology companies such as Voltz, Basecamp Research, Chai Discovery, Dyno, PerturbAI, and Proxima have collaborated with NVIDIA to develop tools to accelerate model-driven treatment design workflows.
Computer-aided drug discovery software providers such as Dassault Systèmes, Cadence (OpenEye), and Schrödinger are integrating toolkit functionality into scientific applications used across discovery teams. Agents can then help generate molecules, dock them, and tailor predictions, turning computer-aided design platforms into systems where researchers can ask questions, initiate analyzes, and identify next-best actions more quickly.
Lab equipment and automation companies such as Automata, HighRes, Tecan, Thermo Fisher, and the autonomous data generation platform Medra are connecting their systems with computational discovery powered by BioNeMo skills.
AI cloud and AI infrastructure companies such as Baseten, Modal, and Nebius use this toolkit to help developers build life science workflows as trusted, hosted services. By supporting BioNeMo’s skills and tools through scalable application programming interfaces, managed computing and production inference environments, these companies can help move agent biology workflows from prototypes to services available to researchers and companies.
availability
The BioNeMo agent toolkit and skills are available on the NVIDIA developer resources page and on GitHub.
