Blueberry Bot: Oregon State University

The Arizona State University Artificial Intelligence Cloud Innovation Center, powered by Amazon Web Services (AWS), partnered with Oregon State University College of Agriculture (OSU) to develop “Blueberry Bot,” a cloud-based solution designed to transform and automate how blueberry growers access expert guidance. OSU Professor, Dr. Wei Yang at the North Willimatte Research and Extension Center, specialized in all aspects of blueberry production systems from field establishment, physiology and nutrition, irrigation, disease and pest management, mechanical harvesting,  to postharvest handling and international marketing. He has over 25 years of experience working with blueberries and developed the first commercial blueberry rootstock for producing blueberry trees. 

By ingesting Dr. Yang’s research and data into an AWS-backed knowledge base and wrapping it in an AI-driven chat interface, Blueberry Bot automated responses to routine cultivation questions, centralizes up-to-date agronomic data, and delivers real-time, personalized recommendations—all without manual intervention. This collaboration accelerates information delivery to growers and provides researchers access to advanced blueberry-specific studies through natural language queries.

Dr. Yang is deeply committed to supporting blueberry growers and strives to respond to their questions and needs as promptly and accurately as possible. However, he receives a high volume of individual inquiries, all of which require his direct attention. This manual process creates a bottleneck that can delay response times and limit his ability to share research findings broadly and efficiently.

Additionally, there is often a natural lag between the completion of research and the publication or communication of results. As a result, the most up-to-date insights may not always be immediately available in responses to growers. The current system also lacks a feedback loop to track common questions and integrate answers into a shared knowledge base, making it harder to streamline support and ensure consistent dissemination of new information.

Blueberry Bot simplifies how expert knowledge is accessed. This chatbot, powered by AWS technology, seamlessly connects interested parties around the globe with Dr. Yang's extensive research and expertise in blueberry cultivation. What sets this solution apart is its commitment to accuracy–it only provides answers when it reaches a 90% confidence threshold. When the bot encounters questions beyond its current knowledge base, it seamlessly transitions to a smart email routing system, connecting growers directly with Dr. Yang for personalized expertise. Behind the scenes, a robust admin interface enables OSU staff to continuously enhance the system's capabilities through performance monitoring, annual knowledge base updates, and detailed analytics. This three-tiered approach, combining intelligent chat interface, automated email handling, and comprehensive admin controls, creates a dynamic knowledge ecosystem that grows smarter with each interaction, making expert agricultural guidance more accessible than ever. 

The following AWS cloud services powered this solution:

• Data Training: Store research PDFs and data in S3, index and score them with Amazon Bedrock, and monitor ingestion with CloudWatch.
• Chat Interface: Host the React.js UI on Amplify, secure logins via Cognito, route queries through API Gateway, and save history in DynamoDB.
• Escalation Workflow: Use Bedrock to flag low-confidence responses, trigger Lambda to collect emails and send them via SES, and log escalations in DynamoDB.write 

By implementing an automated system that handles routine inquiries and manages escalations, researchers like Dr. Yang can be freed from repetitive Q&A tasks to focus on higher-value activities. This creates a powerful multiplier effect where:

Researchers can dedicate more time to:

• Conducting new research and experiments
• Mentoring students and future scientists
• Building valuable industry partnerships
• Writing papers and securing grants

While simultaneously:

• Their existing knowledge reaches more people through the chatbot
• Common questions get instant, accurate responses
• Only truly novel or complex inquiries require their direct attention
• Their responses to escalated questions further enhance the knowledge base

This automation-enabled focus on core research and extension activities ultimately leads to greater scientific impact and faster advancement of the field. The chatbot becomes not just a knowledge sharing tool, but a catalyst for new discoveries by giving experts back their most valuable resource - time.

“When ChatGPT was first released, I was excited about its potential and immediately saw how valuable a similar tool could be for blueberry growers seeking expert guidance on production issues. However, after experimenting with it myself, I realized that a more specialized and robust chatbot was needed to truly meet the unique needs of our industry.

To move this idea forward, I reached out to Tim Rager at OSU Root IT, who played a key role in assembling a talented team of IT professionals from AWS, Arizona State University, and OSU IT. Together, we collaborated to develop what is now the Blueberry Bot.

I found the early logic design of the Chat Bot both amusing and enlightening—it gave me a deeper appreciation of what goes into creating such a tool. I’m grateful to everyone who contributed their time and expertise to bring this vision to life for the blueberry industry.

My hope is that the Blueberry Bot becomes a leading example of how OSU research and Extension faculty can harness AI to expand our collective knowledge and provide timely, impactful support to growers.” 

Wei Yang, Blueberry Specialist and Professor, OSU

“The Blueberry Bot, developed by ASU/AWS, OSU Root IT, and Dr. Yang, who is based at Oregon State University’s North Willamette Research and Extension Center (NWREC), is a prime example of innovation in Extension. It will serve as a starting point in helping increase our Center’s ability to meet the needs of our clientele through cutting edge technology that reaches across generations allowing for even greater innovation and agility.”

Dr. Steve Young, Director, North Willamette Research and Extension Center

"I’ve been working in IT since the early '90s and have handled remote installs with dozens of vendors over the years—this was by far the smoothest experience I’ve seen. The team developing Blueberry Bot walked us through every detail with impressive clarity and professionalism, and the deployment script executed flawlessly. I’m also genuinely impressed with how the bot is architected—modular, intuitive, and with a clean interface that made onboarding easy. Outstanding work all around."  

Curtis Onstott, Infrastructure Manager, OSU College of Agriculture

The framework developed for the Blueberry Bot represents just the beginning of what's possible in AI-powered knowledge dissemination. This open-source solution can be readily adapted across numerous industries, creating specialized AI assistants that bridge the gap between research expertise and practical application. Imagine healthcare professionals accessing the latest medical protocols, legal teams querying vast case law databases, or engineers referencing technical specifications–all through similarly designed chatbots built on trusted research foundations. The key components, including the 90% confidence threshold, expert fallback system, and administrative dashboard, can be replicated to create reliable AI assistants in fields ranging from extension education and environmental sciences to pharmaceutical research and financial services. By following this proven framework, organizations can transform their specialized knowledge bases into interactive, intelligent systems that maintain high accuracy standards while making expert information more accessible to practitioners. The Blueberry Bot demonstrates how research-based chatbots can democratize knowledge sharing across any industry where accessing expert information quickly and accurately is crucial.