Implementing a GenAI-Powered Knowledge Base with AWS

Customer Overview

Our client is a large manufacturer in the heavy building materials industry with operations across the United States and internationally. The organization produces a wide range of construction materials and is part of a broader global network that operates in more than 15 countries and employs approximately 5,500 people.

With a long operating history and a sizable workforce, the company manages an extensive collection of operational documentation, including manuals, diagrams, drawings, tables, and other equipment-related materials. However, due to the vast amount of information stored across multiple systems and locations, employees struggle to quickly locate the specific answers they need, creating inefficiencies in day-to-day operations.

To resolve this challenge, the organization partnered with JBS Dev to develop a streamlined solution that allows users to efficiently search and query their data using natural language.

Challenge

The client faced several key challenges:

• Scattered Information Storage: Important data was spread across various repositories, making access cumbersome and inefficient.
• Inefficient Search Capabilities: Finding specific information required manual searching in different places, leading to delays and inefficiencies in operations.
• Limited Document Accessibility: Users had difficulty accessing the source documents quickly after retrieving information.

The company needed a solution that would allow users to efficiently search a vast amount of data while presenting the results in an easily readable and searchable format. It was also essential for users to be able to ask follow-up questions in natural language and easily access the source documents with a single click.

Solution

JBS Dev designed and implemented a GenAI-powered Knowledge Base using AWS services and Large Language Models (LLMs) to unify and improve the search experience. The solution enables users to submit natural language queries, receive relevant responses, and quickly access the original documents.

Key Features:

• A Web Portal that allows users to:
  • Ask questions using natural language.
  • Upload and categorize additional documents.
  • Manage document metadata.
• A Consolidated Knowledge Base built from Word documents, PDFs, and structured table data.
• Future Scalability: Designed to be able to incorporate images, diagrams, and other unstructured data in future phases.
• Enhanced AI Capabilities: The solution can integrate new LLMs as they become available, ensuring continuous improvement in response accuracy.

Technology Used

JBS Dev leveraged AWS's robust cloud and AI services to build a scalable and efficient platform:

• Data Storage:
  • Documents and structured data were stored in AWS S3 Buckets.
  • The extracted text was processed and chunked based on document types.
• Vector Search and Retrieval:
  • Processed data was stored in AWS OpenSearch, allowing for high-performance retrieval based on semantic similarity.
• Generative AI Processing:
  • AWS Bedrock Generative AI Service hosted the LLM, which retrieved relevant data from OpenSearch and generated user-friendly responses.
• Web Portal Development:
  • Built using Django, a Python-based web framework.
  • Hosted using AWS Lambda, ensuring a serverless, highly available, and fault-tolerant architecture.
  • Integrated with AWS RDS for managing site administration data.
• Security and Access Control:
  • Implemented a full Authorization/Authentication mechanism to ensure secure access.
    
    

Results

The improvement was evident from the moment the solution was deployed. With the new system in place, employees can now ask questions naturally and receive precise answers almost instantly.

An operator, for example, who previously had to search through multiple PDFs and technical manuals to find specifications for a specific piece of equipment can now simply type a query into the web portal. They receive a fast and detailed response, including a summary and a direct link to the original document. This improved approach saved time and reduced frustration and inefficiencies.

Supervisors and managers also found immense value in the solution, as it provided a structured way to access business-critical data without the need for extensive training. The ability to ask follow-up questions in a conversational manner further enhanced user adoption, making the system intuitive and highly effective.

As a result, the company experienced improved productivity, with employees being able to retrieve information very quickly. The solution also laid the groundwork for future enhancements, allowing for additional data types, such as images and diagrams, to be incorporated seamlessly. With AWS's scalable infrastructure and JBS Dev's innovative approach, the company is now better positioned to handle its growing knowledge management needs, ensuring operational efficiency well into the future.

Conclusion

JBS Dev successfully delivered a cutting-edge GenAI solution leveraging AWS services, transforming how the client accesses and utilizes their vast document repository. The system's scalability ensures continuous improvements, positioning the company for long-term success in managing and retrieving critical operational knowledge efficiently.