Product descriptions:
FMC is harnessing the power of science to address agriculture's most significant challenges, protecting crops from the devastating impact of insect pests. And while data is critical to their business, the challenge isn’t big data; it’s having the right data when and where it’s needed. Legacy systems and fragmented data slowed things down, and there was a lack of scalability to build machine learning (ML) solutions that would enable them to work more efficiently. The Databricks Data Intelligence Platform helped unify FMC’s data sources and streamline ML innovation, allowing more effective delivery of timely agronomic insights that help growers manage pests and protect yields.
The need for real-time data to power AI-driven insights
Delivering the right data at the right time is critical for FMC to help growers anticipate and respond to changing conditions — especially when it comes to protecting crops from pests. However, the team faced bottlenecks with legacy systems and architectures, which at times slowed the delivery of insights. FMC’s data scientists were developing promising pest prediction models, but getting those models into production was fragmented. Each team operated independently with its own pipelines and environments, making it difficult to scale results and turn AI insights into real-time data for growers.
“Before Databricks, everyone was working in their own IDE, with little standardization or collaboration. We had to push everything into Azure and manage it ourselves manually,” said Andy Walsh, Data Scientist at FMC.
These challenges came to a head as FMC sought to expand its use of Artificial Intelligence (AI) in the field. Their Arc™ farm intelligence app, which combines local geospatial and weather data with visual inspections of insect pest traps, had the potential to transform how growers manage infestations. Without a unified data foundation, developing and scaling the ML models behind Arc™ farm intelligence would be cumbersome to manage. To unlock the full potential of Arc™ farm intelligence and accelerate innovation in precision agriculture, FMC turned to Databricks.
Accelerating data with ML
With Databricks Data Intelligence Platform, FMC has moved from fragmented, manual processes to a unified environment that drives real-time insights, greater collaboration and faster innovation. Delta Lake and Unity Catalog helped eliminate silos, enforce consistent, fine-grained access controls and ensure reliable data pipelines for ETL and model training alike. FMC utilizes Lakeflow Jobs to enable teams to manage their own workflows and reduce engineering dependence when building pipelines. To further streamline engineering work, they use the feature engineering library in the Databricks feature store. And business users can now explore data independently with Genie, accelerating feedback loops and helping with stakeholder buy-in for AI initiatives.
To accelerate and scale ML, FMC uses Databricks Agent Bricks to build, train and deploy ML models. This has helped FMC transform its Arc pest control app.
“One of our best ML stories is how Databricks enhanced our Arc™ farm intelligence app,” said Michael Seay, Sr. Data Scientist at FMC. “Vision-enabled sensors look at our pest traps and give real-time counts tied to specific crops and locations. We enrich this data to anticipate where infestations are likely to emerge.”
Enriched with geospatial insights from SpatialSQL, the Arc™ farm intelligence app can instantly factor in signals such as location and weather forecast. This has made Arc™ farm intelligence an even more powerful tool for their growers. MLflow automates the entire ML lifecycle from experimentation and versioning to monitoring and performance oversight.
Creating real impact for agriculture
Working with Databricks, FMC has seen measurable gains in productivity and technical performance. FMC has also achieved significant efficiency gains. The number of deployed models has grown from 18 to 33, representing an 83% increase*, while the time to create new models has dropped from weeks to days, resulting in more than 50% fewer person-hours*. This enables the team to scale use cases more quickly and redirect efforts toward higher-value projects.
“Creating a new model went from weeks to days. At times, it could even be within minutes,” said Andy.
Overall, the Databricks Platform has reduced operational complexity, accelerated time to market and enabled FMC to deliver more accurate insights to internal stakeholders and customers in the field. FMC can now share data seamlessly across regions and streamline model development to better enable growers to make informed decisions more quickly.
Looking ahead, FMC is building on its success with the Arc™ farm intelligence app by developing a GenAI chatbot using Databricks Agent Bricks platform. Their goal is to encourage growers to ask questions directly within the app and provide real-time answers and solutions. In parallel, the team is prototyping tabular predictive model outputs using Databricks Apps, enhancing the way data and forecasts are served and visualized. These initiatives enhance FMC’s ability to deliver timely, actionable insights to growers while continuing to scale their data and AI initiatives.
Disclaimer: *Any performance claims/metrics are based on FMC Corporation’s internal analyses and specific use cases, and they are not guarantees.