Customer Overview
Leading Pharma company is a premier contract research, development, and manufacturing organization (CRDMO) that delivers comprehensive, integrated discovery, development, and manufacturing services to a diverse clientele. Serving the pharmaceutical, biotechnology, animal healthcare, consumer goods, and agrochemical sectors, Customer partners with industry leaders, including global multinationals, SMEs, non-profits, academic institutions, and government organizations.
With a team of over 6,000 scientists, Customer combines unparalleled expertise with substantial capacity to deliver exceptional science. Commitment to robust data security, high-quality manufacturing, and rapid execution helps clients accelerate time-to-market and reduce the cost of innovation. These capabilities ensure Customer is deeply involved in pioneering advancements across various industries.
Business Need
Customer faced a significant challenge with their bulk collection of contract and legal documents stored in their on-premises environment. Retrieving specific content from this extensive document base was cumbersome, primarily due to the prevalence of scanned image copies, which were not in digital text format. This reliance on non-digital documents necessitated a labour-intensive, manual process for document retrieval, significantly hampering efficiency and productivity.
Customer’ s extensive collection of over 7,500 contract documents presents several technical challenges and inefficiencies in their processes:
- Lengthy Documents: Each contract spans over 40 pages, making it difficult to quickly locate and extract relevant information manually.
- Multiple and Diverse Contract Types: The presence of various contract types (commercial, non-commercial, hybrid) along with more than 12 different document types necessitates distinct data extraction methodologies, complicating the identification and retrieval process, and making manual extraction highly prone to errors and inconsistencies.
- Manual Analysis: The manual review and analysis of these documents are not only labour-intensive but also increase the risk of human error, potentially leading to critical information being overlooked or misinterpreted.
- Time and Resource Drain: The current manual processes consume a significant amount of time and resources, drastically reducing overall operational efficiency and potentially delaying key business operations.
To address these challenges, Customer planned to implement a GenAI-powered Intelligent Document Management System (IDMS) to enhance their document processing capabilities technically.
Quadra at Work and solution design
Understanding customer’s pain points and business needs, Quadra as an advanced consulting partner, collaborated and designed a GenAI powered solution for their requirement which includes,
- Amazon Textract: Utilized to extract data from scanned copies of Customer’s documents, converting them into digitalized text format which employed machine learning to accurately recognize text and structured data within scanned images, facilitating downstream processing.
- Amazon Bedrock and Bedrock Guardrails: Leveraged serverless Large Language Models (LLMs) and FMs – Amazon Titan Premier and Amazon Nova Pro – for advanced document processing and user prompts which ensured the secure deployment and operation of these LLMs and FMs with guardrails configured for providing safe and efficient AI-driven document analysis.
- Amazon S3: Centralized cloud-based storage repositories for securely storing all documents, both the original scanned copies and their digitized versions.
- AWS Lambda: Serverless functions designed to orchestrate and automate the entire GenAI workflow. AWS Lambda is used to invoke document processing functions, including calling Amazon Textract for OCR, interacting with LLMs through Amazon Bedrock, and saving processed outputs back to Amazon S3.
- Amazon API Gateway: Provided a secure and scalable entry point for applications to interact with backend Lambda functions, facilitating the integration of GenAI processing capabilities into Customer’s Intelligent Document Management System.
- VPC Private Link: Utilized to establish secure and private connectivity between Customer's network and AWS services. By leveraging VPC Private Link, Customer ensured that all traffic to AWS services remains within the AWS network, minimizing exposure to the public internet and enhancing security.
- Amazon CloudWatch: Utilized for comprehensive resource monitoring, including metrics, logs, dashboards, and alarms.
- Amazon CloudTrail: Implemented for tracking API call events.
Benefitted Outcomes
- Managed FMs: Leveraged Amazon Bedrock to harness the power of Foundation Models (FMs). benefiting from usage-based pricing and seamless API integrations. This resulted in a 40% reduction in operational costs and 60% faster deployment times for AI-powered applications.
- Amazon Bedrock Guardrails: Implemented to ensure safe and responsible AI applications by filtering harmful content, blocking undesirable topics, and preventing hallucinations in model responses. This led to a 40% decrease in content moderation incidents and a 45% improvement in user trust and satisfaction.
- Increased Efficiency in Document Retrieval: Average retrieval time decreased from days to minutes, resulting in a 96% reduction in retrieval time.
- Enhanced Accuracy in Data Extraction and Categorization: Significantly reduced document processing and categorization errors; error rates dropped from 15% to less than 1%, achieving over 93% improvement in accuracy.
- Infrastructure-less Design: The solution was built using AWS native and serverless architecture, eliminating the need for infrastructure management and optimizing the infrastructure team's efficiency by 80%, allowing them to focus on innovation.
