Global data doesn’t always solve local problems. That’s why Factored is partnering with NASA Harvest, Arizona State University, and MLCommons on the Grow by Numbers Challenge, a collaborative effort to rethink how we train geospatial AI.
The Challenge: When Global Data Falls Short Locally
Satellite imagery has revolutionized how we understand agriculture, climate, and disasters. But here’s the catch: training AI models on global data doesn’t guarantee good performance in specific regions. That’s where the Grow by Numbers Challenge comes in.
Launched by NASA Harvest, this international challenge is funded through a grant awarded to Arizona State University. It is supported by MLCommons, with technical contributions from Factored:
How do we select the right training data to build better, more localized AI models for agriculture?
Factored’s Role: Building the Backbone for Smarter AI
Factored is driving the technical engine behind the Grow by Numbers Challenge. Through our Machine Learning Center of Excellence, our engineers co-develop the benchmarking systems that enable smarter, fairer AI, built to solve real-world problems at scale.
Our contributions include:
- Scalable evaluation pipelines
- Reliable performance testing across diverse regions
- Data-centric solutions for satellite imagery challenges
Why It Matters
The impact of this challenge goes beyond crop classification. It sets the foundation for:
- Accurate poverty and disaster mapping
- Improved food security monitoring
- Scalable AI applications across low-resource regions
By empowering researchers to select the most useful samples from the CropHarvest dataset, we’re helping machines learn how to see and serve the world better.
Meet the People Behind the Tech
This initiative is powered by global collaboration and anchored by local insight. Factored’s engineers, including Rafael Mosquera & Sara Hincapié from our AI & ML Center of Excellence, ensure the technology driving this challenge is not only fast but fair and future-ready.
Join the Movement
The Grow by Numbers Challenge runs until August 3, 2025, and is hosted on DynaBench. It’s open to researchers, students, and ML enthusiasts everywhere.
Learn more and join the challenge
