Agriculture Trial Explorer

Unable to compel a large group of scientists to conform to standards when designing protocols?

Now you don't need to.

Initial Situation/Problem statement

For large multi-national organizations, the need for standardized documentation is equally immense as it is unachievable. Take our customer Syngenta for example. With over 30,000 employees, working in over 90 countries, the diversity of cultures and customs is staggering. In an environment like this, achieving a standardization in even something as important as their Agricultural testing Protocols. The document that dictates how and agricultural trial is conducted and its results measured. Becomes a herculean task. This is where we come in.

Project Goal

Our goal with this project was to leverage large language models (LLMs), in order to achieve parity in the diversity of protocols. No longer would each researcher use their own methodology, and nomenclature. Instead, all protocols that were used would be standardized to the highest possible extent, without disrupting the workflows of the existing researchers, and risking a loss of productivity.

Developed Solution and its benefit

Throughout the project, the team explored several vastly different strategies to achieve this goal.

In the end, in order to optimally achieve our goals with the resources provided, we decided upon a two-stage approach.

First, the generated protocols are reviewed by an LLM, identifying key issues and areas that differ strongly from the desired structure. The result of this review is then passed into our solution, which uses this information to generate an entirely new protocol.

This approach has two major advantages. Firstly, it does not disrupt the customer’s existing workflow; researchers continue working exactly as they did previously. The second is that through the use of LLMs, the language is standardized. Complex sentence structures are simplified, and gaps in language quality are improved.

Due to copyright restrictions we cannot show any further screenshots of the application.

Testimonial
«Working with the student team on the Protocol Assistant project was a very positive experience. They quickly understood the complexity of research protocols and developed a structured, intelligent tool that supports the review and validation process. Their combination of technical skill and curiosity led to a functional prototype that lays the foundation for automating and standardising protocol management. The collaboration was smooth, professional, and delivered clear value for our ongoing work.»
Mladen Cucak - Data Scientist

Key terms

• Programming languagesJava, Javascript
• Frontend StackReact.js
• Backend StackQuarkus
• DatabaseHibernate
• Deployment PlatformAWS/Switch

Customer

Syngenta Crop Protection AG
Mladen Cucak
Rosentalstr. 67
4058 Basel, Switzerland
www.syngenta.com

Team