Accelerating Drug Discovery Pipelines With AI
Development Platform for AI-Driven Biology and Drug Discovery
NVIDIA Clara
Build With Open Source AI Models and Datasets
NVIDIA Clara™ is a family of open-source AI foundation models, tools, and recipes for biomedical research. Clara includes AI foundation models for omics, protein and molecule structures, imaging, 3D anatomy, surgical robotics, and digital twins for physics-informed simulations. They’re built using leading open clinical, imaging, and biological datasets.
Clara integrates directly with the NVIDIA AI software stack. It’s complete with pretrained checkpoints and novel model architectures, training recipes, curated open datasets and tooling, and unified benchmarking and evaluation frameworks.
NVIDIA BioNeMo Framework and Recipes
Train and Customize Models for BioPharma
NVIDIA BioNeMo™ Framework is an open source machine learning framework for building and training deep learning models for biopharma.
By accelerating the most expensive and prolonged aspects of AI model development, AI model builders doing biomolecular research with DNA, RNA, and protein data can access tools to scale their research to new heights.
The platform includes curated training recipes, data loaders, and pretrained, optimized AI model architecture examples that are domain-specific and accelerated for the best performance, making AI model building faster and simpler.
NVIDIA AI Blueprints
Build Generative AI Applications for Drug Discovery With Pretrained AI Workflows
NVIDIA BioNeMo Blueprints are pretrained reference workflows designed for generative AI applications in drug discovery, offering biopharma teams a foundation to accelerate research and innovation.
For organizations in biopharma looking to integrate AI, these blueprints include reference code, tools, and comprehensive documentation to customize and deploy workflows tailored to proprietary data and unique therapeutic pipelines.
These resultant data flywheels enhance model performance over time, driving deeper insights and accelerating discovery. BioNeMo Blueprints enable biopharma teams to streamline innovation and unlock the potential of AI in drug development.
NVIDIA NIM Microservices
Optimized AI Inference for Scalable Drug Discovery Applications
NVIDIA BioNeMo NIM™ includes a set of optimized, easy-to-use AI microservices that enable gigascale inference and new capabilities in drug design.
Built for computational drug discovery platform developers and data scientists, NIM microservices are constructed as containers, providing everything needed for the most efficient, portable deployment available, including easy API integration into enterprise-grade AI applications.
NIM microservices increase the efficiency and innovation of drug discovery workflows while decreasing the total cost of ownership and time to market for AI-driven discoveries.
NVIDIA CUDA-X for Biopharma
GPU-Accelerated Tools and Libraries for Molecular AI
NVIDIA CUDA-X™ libraries deliver drop-in code modules that accelerate the most compute-intensive layers inside today’s biomolecular AI models so researchers can innovate faster.
cuEquivariance is an NVIDIA Python library designed to facilitate the construction of high-performance equivariant neural networks using segmented tensor products. cuEquivariance also has optimized kernels for triangle attention and triangle multiplication for modeling pairwise interactions in protein structure prediction (e.g., AlphaFold-style architectures).
One-line code changes and out-of-the-box integrations through PyTorch and Jax bindings make it easier than ever to swap out the most expensive steps in a workflow with CUDA®-optimized kernels.
cuEquivariance unlocks new performance levels in protein-structure, generative-chemistry, and molecular-dynamics models.
Additional Technologies
Get Started
Documentation
Find a collection of documents, guides, manuals, how-tos, and more in the NVIDIA BioNeMo Documentation Hub.
Product Updates
Get notified of new releases, bug fixes, critical security updates, and more for biopharma.
Contact Us
Got questions? Our healthcare experts are here to assist you!
FAQ
NVIDIA BioNeMo is an open development platform for AI-driven biology and drug discovery. It brings together open models, tools, and datasets across the AI lifecycle, enabling researchers and developers to build, customize, and deploy AI applications for biological research and drug discovery. BioNeMo enables lab-in-the-loop workflows that close the feedback loop between experimentation and AI, using experimental data to continuously improve model intelligence, predictions, and scientific insight.
- NVIDIA BioNeMo Framework is a collection of programming tools and libraries that streamline the building of biomolecular AI models, accelerating the most time-consuming and costly stages of computational drug discovery.
- NVIDIA NIM is a set of optimized, easy-to-use, portable microservices designed for biomolecular scientists in drug discovery needing secure, reliable AI model inferencing.
NVIDIA BioNeMo Framework: Users can access BioNeMo Framework in two ways. NVIDIA’s offering for enterprise-grade use of BioNeMo with an NVIDIA AI Enterprise license offers the BioNeMo container via the NVIDIA NGC™ catalog, which provides enterprise developers and researchers with a secure, scalable toolchain to build biomolecular workflows. The open source version of BioNeMo that researchers and data scientists use is available for installation from GitHub, including all of its components.
Pretrained Models: BioNeMo Framework offers BioNeMo recipes that show example implementation of Transformer Engine-compatible architecture and also provides for easy testing and fine-tuning. Users are encouraged to extend their own implementations using BioNeMo recipes as tutorials.
BioNeMo NIM microservices offer various optimized checkpoints such as Evo2 (genomic generative AI), GenMol (molecule generation), DiffDock (docking), and others, each supporting their respective biomolecular tasks.
See the latest list of system requirements for BioNeMo Framework on the NGC Catalog Container page.
See the latest system requirements for NIM microservices in NVIDIA API Documentation.
BioNeMo Framework code is licensed under Apache 2.0, while the official NGC containers are governed by the NVIDIA AI Product Agreement (and can be covered by an NVIDIA AI Enterprise license for commercial use).
NVIDIA BioNeMo exposes its models through libraries, APIs, and containerized NIM microservices, allowing teams to call structure‑prediction, generative‑chemistry, and other functions directly from their established drug‑discovery pipelines.
Launch the BioNeMo Framework container or clone the GitHub repository, point a YAML config to your data, add a restore_from_path (or equivalent) to load existing weights, and run the provided training script to pretrain or fine‑tune a model.
Because BioNeMo Framework is hosted openly on GitHub, developers can fork the repository and submit pull requests under the “CONTRIBUTING” guidelines of the repository documentation to add new code.
Please contact us with your question.
Additionally, users can access documentation, tutorials, and an active developer forum via the BioNeMo Resources Hub, while enterprise customers can open support tickets through NVIDIA AI Enterprise.