Business Objective / Goal
To enhance regulatory adherence, document clarity, and operational efficiency by implementing an AI-based web platform for SOP evaluation and rewriting—addressing variability, complex language, and manual review bottlenecks in Biocon’s document processes.
Solutions & Implementation
- Developed a web-based Generative AI application using ReactJS (frontend) and Django (backend) hosted on AWS EC2.
- Integrated Amazon S3 for document storage and Amazon RDS (PostgreSQL) for prompt templates and revision history.
- Used prompt engineering to transform SOPs with LLM responses via Amazon Bedrock (Claude 3 Sonnet).
- Supported manual import of revision inputs (from emails or MS Teams) into a versioned revision tracker.
- Enabled tracked changes, structural rewriting, grammar checks, and export-ready outputs aligned to DMS formats.
- Ensured CFR audit readiness through version control, traceability, and compliance reporting.
Major Technologies Used
- Amazon Bedrock (Claude 3 Sonnet) – LLM for rewriting SOPs
- Amazon S3, EC2, RDS (PostgreSQL), AWS Textract, Amazon MQ – Scalable AWS infrastructure
- ReactJS, Django, Python – Application logic and front-end
- LangSmith, Pinecone – Prompt engineering and semantic vector storage
- AWS Athena, Presto – For future extensibility of SOP data querying
Business Outcomes
- Streamlined SOP Rewriting and Review Workflow Enabled structured, tracked rewriting and feedback within a single AI interface, reducing manual document iterations.
- Improved SOP Quality and Readability Rewrote SOPs using consistent tone, improved grammar, and standardized structure across all operational sites.
- Audit-Ready Change Traceability Integrated revision tracking, version control, and compliance feedback mechanisms aligned to CFR standards.
- Foundation for AI-Enabled Regulatory Compliance at Scale Built a scalable GenAI-enabled platform ready for Phase 2 features like FDA guideline validation, automated compliance scoring, and integration with Biocon’s SOP DMS.
