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 scale autonomous chemical-free weeding.
Reported impact
Aigen trained production weed-control models with 99 percent synthetic data, reduced onboard compute utilization to about 30 percent, shrank onboard hardware from five compute devices to three Jetson modules, and compressed model turnaround from seasons to hours so new crops can be deployed in the same season.
Aigen built Alchemy as an outdoor world model on NVIDIA Cosmos to replace slow field-data collection with synthetic training generation for agricultural robotics. The company post-trained NVIDIA Cosmos 3, and earlier Cosmos Transfer 2.5, on fleet-captured RGB and depth clips so the model could generate realistic variations in crops, weeds, soil, lighting, and depth for perception training. That synthetic-data pipeline lets Aigen train field-weeding models with only 1 percent real-world data while expanding coverage to scenarios the team has not physically collected. On the robots themselves, Aigen uses three NVIDIA Jetson Orin NX modules for perception, navigation, and mechanical control, reducing onboard compute utilization from roughly 100 percent to about 30 percent and staying within a solar-power budget. This pairing of synthetic-data generation and efficient edge AI gives Aigen a path to scale autonomous, herbicide-free weed control across more crops and field conditions.
Fleet-captured RGB and depth clips seed Alchemy and Cosmos training, the world model generates labeled synthetic crop and weed scenarios, perception models are trained from that data, and Jetson Orin NX modules run the resulting perception and control stack onboard Aigen's solar-powered field robots.
"Normally that takes three to four years of collecting data, labeling it, and training models ... model turnaround is hours instead of seasons."
Aigen built Alchemy, an outdoor world model that generates realistic synthetic variations in crops, weeds, soil, lighting, and depth.
Aigen's newest robots use three NVIDIA Jetson Orin NX modules for energy-efficient perception, navigation, and mechanical control.
During development testing, Aigen reports that the Jetson-based architecture operated at approximately 30% compute utilization.
Using synthetic training data from Alchemy, Aigen achieved autonomous field weeding with a perception model trained on 1% real-world data and 99% synthetic data.
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.
Related
NVIDIA Sample Sep 30, 2026 Corteva 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… 15-fold BioNeMo Boltz-2 NIM · NVIDIA H200 GPUs · NVIDIA cuML 
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 