NVIDIA Clara

Type: Platform Tags: NVIDIA, healthcare, medical imaging, genomics, drug discovery, AI, radiology, clinical AI, life sciences Related: NVIDIA-Parabricks, NVIDIA-Clara-Viz, NVIDIA-MONAI-Toolkit, NVIDIA-FLARE, NIM-for-MAISI, NIM-for-VISTA-3D, NVIDIA-BioNeMo, BioNeMo-Recipes, NIM-for-AlphaFold2, NIM-for-OpenFold3, NIM-for-Boltz2, NIM-for-Evo-2, NIM-for-MSA-Search, NIM-for-ProteinMPNN, NIM-for-RFdiffusion, NIM-for-MolMIM, NIM-for-GenMol, NIM-for-DiffDock, NIM-for-ALCHEMI-Batched-Geometry-Relaxation, NIM-for-ALCHEMI-Batched-Molecular-Dynamics, NVIDIA-Holoscan, NVIDIA-Riva, NVIDIA-AI-Enterprise, NVIDIA-NIM, NGC, TensorRT Sources: NVIDIA official documentation; https://docs.nvidia.com/clara/index.html, https://docs.nvidia.com/bionemo-framework/latest/main/recipes/, https://docs.nvidia.com/nim/medical/maisi/latest/overview.html, https://docs.nvidia.com/nim/medical/vista3d/latest/overview.html, https://docs.nvidia.com/nim/bionemo/msa-search/latest/overview.html, https://docs.nvidia.com/nim/bionemo/proteinmpnn/latest/overview.html, https://docs.nvidia.com/nim/bionemo/rfdiffusion/latest/overview.html, https://docs.nvidia.com/nim/bionemo/molmim/latest/overview.html, https://docs.nvidia.com/nim/bionemo/genmol/latest/overview.html, https://docs.nvidia.com/nim/bionemo/diffdock/latest/overview.html Last Updated: 2026-04-30

Summary

NVIDIA Clara is a healthcare AI computing platform that provides domain-specific frameworks, tools, and reference applications for medical imaging, genomics, drug discovery, and clinical AI. It is organized into several focused sub-platforms — Clara Imaging, Clara Parabricks (genomics), and Clara Holoscan (medical-grade edge AI) — each offering GPU-accelerated pipelines and pre-trained models tailored to clinical and life sciences workflows. Clara enables hospitals, medical device companies, and pharmaceutical researchers to develop, validate, and deploy AI-powered diagnostics and therapeutics.

Detail

Purpose

Healthcare AI development faces unique challenges: specialized data types (DICOM, FASTQ, whole-slide images), regulatory requirements (FDA 510(k), CE marking), de-identification requirements, scarce labeled datasets, and the need for explainability. Clara addresses these with domain-specific frameworks that handle medical data ingestion, federated learning for privacy-preserving model training, pre-trained models for common medical imaging tasks, and real-time inference pipelines for clinical deployment. It bridges the gap between GPU computing capabilities and healthcare domain requirements.

Key Features

Clara Imaging (Medical Imaging AI):

• NVIDIA-MONAI-Toolkit: NVIDIA AI Enterprise-supported MONAI development sandbox for medical imaging AI, including MONAI Core, MONAI Label, NVIDIA FLARE integration, and curated pretrained models.
• Medical imaging NIMs: NIM-for-MAISI generates synthetic 3D CT images and annotation masks for research workflows, while NIM-for-VISTA-3D provides interactive 3D segmentation and annotation.
• MONAI Deploy: MLOps framework for packaging, validating, and deploying medical imaging AI as MONAI Application Packages (MAPs) integrated with hospital PACS/VNA systems
• Pre-trained Segmentation Models: AI models for organ segmentation (liver, lungs, kidneys, brain), lesion detection (lung nodules, liver tumors), and annotation (total body segmentation)
• NVIDIA-Clara-Viz: CUDA-based 2D/3D medical image visualization and digital pathology viewing.

NVIDIA-Parabricks (Genomics):

• GPU-Accelerated Variant Calling: Full GATK-compatible secondary analysis pipeline (alignment, duplicate marking, variant calling) running in 20–50 minutes on GPU vs 24–48 hours on CPU
• FASTA-to-VCF Pipeline: End-to-end germline and somatic variant calling from raw sequencing reads
• Multi-GPU Scaling: Linear scaling across multiple GPUs for large cohort studies
• Compatibility: Output VCF files are bit-for-bit compatible with CPU-based GATK, ensuring regulatory compliance and interoperability

Clara Holoscan (Real-Time Medical AI):

• See NVIDIA-Holoscan — Clara Holoscan is now its own branded platform for streaming AI sensor processing in medical devices

Use Cases

• Radiology AI: automated organ and lesion segmentation for radiologist workflow augmentation
• Pathology AI: whole-slide image analysis for cancer grading and tumor detection
• Genomics: rapid population-scale variant calling for oncology, rare disease, and GWAS studies
• Federated learning: multi-site model training across hospital networks without sharing patient data using NVIDIA-FLARE
• Surgical AI: real-time instrument detection and procedure guidance in the OR (Holoscan)
• Drug target identification and structure prediction for pharmaceutical R&D (BioNeMo integration)

Hardware Requirements / Compatibility

• Clara Parabricks: NVIDIA V100, A100, H100 recommended; minimum GTX 1080 or T4; CUDA 11.x+; Linux only
• MONAI/Clara Imaging: Any NVIDIA GPU with CUDA 11.x+; commonly deployed on A100, H100; Docker containerized
• Storage: Large-scale genomics pipelines require high-throughput NVMe storage (GPUDirect Storage recommended for large cohorts)
• OS: Ubuntu 18.04/20.04/22.04; Red Hat Enterprise Linux 7/8

Language Bindings / APIs

• Python: MONAI Python library (pip install monai); Clara Parabricks CLI (pbrun command)
• C++: Holoscan SDK C++ API for real-time sensor processing
• CLI: pbrun for Parabricks; clara CLI for model registry and pipeline management
• DICOM/HL7 FHIR: MONAI Deploy integrates with hospital PACS via DICOMweb and HL7 FHIR standards
• REST API: MONAI Label server REST API for active learning annotation workflows

Connections

• NVIDIA-Parabricks — genomics acceleration component of Clara for next-generation sequencing workflows.
• NVIDIA-Clara-Viz — Clara visualization component for 2D/3D medical imaging and pathology.
• NVIDIA-MONAI-Toolkit — enterprise-supported MONAI distribution for medical imaging AI development.
• NVIDIA-FLARE — federated learning and federated analytics runtime for privacy-preserving multi-party healthcare AI.
• NIM-for-MAISI and NIM-for-VISTA-3D — current medical imaging NIMs for synthetic CT generation and 3D segmentation/annotation.
• NVIDIA-BioNeMo and BioNeMo-Recipes — BioNeMo is the drug discovery/computational biology layer of Clara; recipes cover the current public training-reference implementation layer.
• NIM-for-AlphaFold2, NIM-for-OpenFold3, NIM-for-Boltz2, and NIM-for-Evo-2 — representative BioNeMo NIMs for structure, interaction, and DNA sequence workflows.
• NIM-for-MSA-Search, NIM-for-ProteinMPNN, NIM-for-RFdiffusion, NIM-for-MolMIM, NIM-for-GenMol, and NIM-for-DiffDock — additional BioNeMo NIMs for sequence search, protein design, molecular generation, and docking.
• NIM-for-ALCHEMI-Batched-Geometry-Relaxation and NIM-for-ALCHEMI-Batched-Molecular-Dynamics — ALCHEMI NIMs for atomistic modeling adjacent to drug discovery and materials workflows.
• NVIDIA-Holoscan — Clara Holoscan became the NVIDIA Holoscan SDK; the real-time streaming AI platform for medical devices
• NVIDIA-Riva — Riva provides speech AI for clinical documentation (ambient clinical intelligence use cases)
• NVIDIA-AI-Enterprise — Clara workloads run under AI Enterprise license for regulated healthcare environments
• NVIDIA-NIM - medical imaging, BioNeMo, and speech healthcare capabilities can be deployed as NIM microservices.
• NGC — Clara containers, Parabricks, MONAI models, and reference applications distributed via NGC
• TensorRT — TensorRT accelerates inference for MONAI segmentation models and Clara imaging pipelines

Resources

• NVIDIA Clara Platform
• MONAI Framework
• MONAI Deploy
• Clara Parabricks Documentation
• Clara on NGC