Case Study – Generative AI

Renowned Healthcare Data Entry Provider- Intelligent document processing

About the customer

The company is a premier provider of healthcare data entry services, specializing in the extraction and organization of data from diverse sources and formats. Utilizing state-of-the-art technologies, including generative AI, Chatmate AI excels in the efficient collection, transformation, and storage of healthcare data. This advanced approach ensures streamlined processes and enhanced data management for healthcare institutions, enabling them to focus on delivering superior patient care and operational efficiency.

Challenges

Use Case: GENERATIVE AI – using Amazon Bedrock

The company is into healthcare data entry providing services that specializes in collecting data from various sources in different formats. By leveraging cutting-edge technologies, such as generative AI, they efficiently collect, transform, and store this data in an organized form, enabling streamlined processes and improved data management for healthcare institutions.

The Major Challenges:

  1. Manually searching through documents for specific information was time-consuming and inefficient. This led to delays and potential errors in data retrieval.
  2. As the volume of documents grew significantly, managing and retrieving information became increasingly challenging, leading to inefficiencies and delays in accessing crucial data.
  3. The existing system struggled to scale effectively with the growing volume of documents, causing bottlenecks and reduced performance as demands increased.
  4. Coordinating document information retrieval with other systems or processes was difficult, resulting in fragmented workflows and increased chances of miscommunication.
  5. Significant human resources were required to manage and operate the process, increasing operational costs and diverting valuable resources from other critical tasks.
thin

Solution

Operisoft proposed a targeted solution to effectively address the challenges faced by customers by recommending the use of AWS Generative AI services, which offer advanced capabilities for document processing and data extraction. The solution will utilize AWS Textract to analyze and process documents, extracting specific key-value pairs that will be securely stored in an output bucket, accessible via a dedicated API. This approach ensures a powerful and efficient system that leverages AWS’s AI and machine learning technologies to deliver precise and timely responses. The proposed solution is designed to meet the customer’s needs by providing a robust, scalable, and secure system for extracting essential values from documents.

LLM Customization:

Team Operisoft customized an LLM for a RAG application by cleaning and pre-processing text data, structuring non-text data, and chunking it for efficient retrieval. Using embedding models, the data is converted into high-dimensional vectors, indexed for fast similarity searches. A retrieval chain is set up to query the indexed documents and provide precise answers, with prompt templates ensuring structured outputs like JSON. Multiple tasks are processed in parallel for efficient information extraction, gathering all results in one place for easy access and analysis.

thin

Solution Diagram

AWS Services used:       

  • AWS EC2
  • Lambda functions
  • API Gateway
  • S3


  • Cloudformation
  • Amazon Cognito
  • Amazon Textract
  • Systems Manager Parameter Store
  • Amazon Bedrock


Outcome


  1. Leveraging AWS Generative AI services, Operisoft’s solution ensures advanced document processing capabilities, significantly reducing the time required to analyze and extract data by up to 70%.
  2. With the use of AWS Textract, the solution achieves precise extraction of key-value pairs from documents, enhancing the accuracy of the data retrieved by up to 90%.
  3. Extracted data is securely stored in an output bucket, ensuring 100% data integrity and compliance with security standards.
  4. The dedicated API provides easy access to the stored data, enabling seamless integration with other systems and facilitating up to 50% faster retrieval of essential information.
thin