Logo
AI-Powered Knowledge Platform

Centralizing Almond Industry Knowledge with an AI-Powered RAG Chatbot

A leading California almond company partnered with Waverley Software to build an AI-powered chatbot that consolidates fragmented industry knowledge into a single, source-ranked interface for growers across web and mobile
Background

This company is a leading California-based almond company

"with deep roots in the Central Valley, which approached Waverley Software with a clear challenge."

Authoritative knowledge about almond cultivation, pest management, soil science, and industry best practices was scattered across dozens of sources, UC research papers, internally gathered content, board publications, and other internal materials.

As a result, growers and industry professionals had no single, intelligent interface to access this information quickly and reliably.

The Company’s goal was to build an AI-powered chatbot that could consolidate this knowledge, deliver source-ranked answers in real time, and be accessible on both web and mobile platforms, serving as a trusted resource for the entire almond growing community.

Phase 1:

Foundation and Proof of Concept

During this phase, the team designed and implemented the core RAG (Retrieval-Augmented Generation) backend, including a document ingestion pipeline capable of processing PDFs, the ability to add web-scraped content, OpenAI embedding generation, and a vector database for semantic retrieval. A source prioritization logic was established, ranking Almond Doctor content first, followed by UC research, Board publications, materials, and other sources, ensuring users always received the most authoritative answers available.

The PoC validated that the system could reliably answer domain-specific agricultural questions with source references, with response times under two seconds under normal load.

Phase 2: Full MVP Delivery

The project started with a clear understanding

With the PoC confirming the approach, the team moved into full MVP development. This phase delivered a complete, production-deployed system across three surfaces: a responsive web application, all featuring a ChatGPT-style conversational interface with multi-turn Q&A, image upload support, and visible source references in every response. An administrative interface was built in parallel, giving RPAC staff the ability to upload and manage documents, configure source priorities, monitor ingestion status, and search all RAG-indexed materials, without requiring engineering involvement in day-to-day content operations. The entire solution was deployed on Azure with a full CI/CD pipeline, logging, monitoring, and security hardening in place.

Centralized industry knowledge:

For the first time, growers and professionals can access consolidated, source-ranked answers from across the almond knowledge ecosystem through a single conversational interface

Multi-platform reach:

The simultaneous delivery of web, iOS, and Android applications ensures the chatbot is accessible in the field, not just at a desk. It was challenging to maintain a real-time understanding of inventory health due to fragmented, infrequently updated data

Image-in-context Q&A:

Users can upload photos of crops, pests, or soil conditions directly within a conversation, enabling more accurate and contextual responses

Automated content pipeline:

The document ingestion system handles PDF extraction and web scraping automatically, reducing the manual effort required to keep the knowledge base current

Admin self-sufficiency:

The administrative interface empowers the RPAC team to manage content, priorities, and sources independently, removing ongoing dependency on the development team

RAG settings:

Configuration of the whole answering pipeline, including model parameters (prompt, temperature, etc.) and knowledge base

Business Delivered Value

This engagement delivered measurable value to the customer and the broader community it serves:

More reliable access

to authoritative agricultural knowledge for growers and industry professionals

Significant reduction in time

Spent searching across fragmented sources

A scalable production-ready platform

that can grow with new content, new sources, and new user needs

Full operational independence

for the RPAC team through the administrative interface

A strong Technical foundation

cloud infrastructure, CI/CD, monitoring — for future feature development and expansion

Our Main Goal?

The goal was to consolidate fragmented industry knowledge into a single AI-powered interface, giving growers and professionals real-time, source-ranked answers across web and mobile

Frequently asked questions

Ready to turn scattered expertise into instant, trusted answers?

Unify fragmented industry knowledge into an AI-powered chatbot growers can trust