Corteva needed to evaluate billions of host-pathogen protein interactions and massive metabolomics datasets faster, but early-stage agricultural discovery pipelines were too computationally expensive to prioritize hypotheses quickly or refresh natural-product analyses often enough.
Reported impact
Corteva cut host-pathogen interaction mapping from years to weeks, made BioNeMo Boltz-2 NIM 4.15x faster than the open-source baseline, accelerated mass-spectral clustering 80x to 100x, and moved metabolomics runs from quarterly to monthly so more candidates can be evaluated earlier.
Corteva built two accelerated discovery workflows on NVIDIA software rather than treating crop research as one monolithic HPC job. For seed innovation, BioNeMo Boltz-2 NIM running on H200 GPUs predicts and verifies host-pathogen interactions, allowing the team to map disease-related protein networks in weeks instead of years while keeping biological fidelity. For crop protection, Corteva and the Wang Bioinformatics Lab rebuilt metabolomics analysis around RAPIDS cuML and cuVS so mass-spectral clustering, embeddings, and molecular networking can run at much higher cadence across growing natural-product libraries. Those two paths share a common pattern: proprietary biological data is filtered and prioritized on accelerated infrastructure first, then researchers use the resulting interaction maps and molecular networks to decide which candidates deserve scarce wet-lab and field resources downstream.
Pathogen and host-protein data are filtered and passed through BioNeMo Boltz-2 NIM on H200 GPUs to produce interaction maps for seed research. Separately, mass-spectrometry data enters a cuML and cuVS pipeline that clusters, embeds, and networks metabolites so researchers can prioritize natural-product candidates for crop-protection follow-up.
BioNeMo Boltz-2 NIM is 4.15x faster than the open-source model.
Enabled by the acceleration, the Corteva team can now run the workflow more frequently, from quarterly to monthly.
NVIDIA cuML and cuVS enable 80x to 100x speedup in mass spectral clustering and a 50x to 60x speedup in embedding.
This increase in speed and scaling enables mapping the protein interaction network in weeks instead of years.
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
The official contact page lists "Corporate Headquarters" at "9330 Zionsville Road, Indianapolis, IN 46268," supporting a United States headquarters country.
Related
NVIDIA Sample Sep 4, 2026 Aigen Aigen needed to train field-robot perception and control systems for highly variable agricultural environments, but collecting and labeling enough real-world crop and weed data was too slow, seasonal, and expensive to… 99% synthetic data and 1% real-world data Alchemy · Cosmos Transfer 2.5 · NVIDIA Cosmos 3 
Google Cloud Sample Modified · Oct 9, 2026 Luxia Agro Luxia Agro is an Argentine company specializing in the import of phytosanitary products for the country's agricultural sector. With a core team of just 10 people, the company has built a reputation for quality and… 50% reduction in labor hours Cloud Run · Cloud SQL · Cloud Storage NVIDIA Sample Oct 9, 2026 California Department of Transportation (Caltrans) Caltrans needed to inspect and monitor roughly 1,200 active transportation projects more frequently and safely, but manual site visits were slow, inconsistent, and risky for staff working around traffic, bridges, and… Remote operations up to 100 miles Esri ArcGIS · NVIDIA CUDA · NVIDIA Jetson Orin 