• Automated Collibra catalog upkeep with an agentic RAG system that performs acronym discovery and generates plain-language logical names, reducing a recurring 8-hour manual task to 30 minutes.
• Delivered a terminology RAG that batches acronym/abbreviation resolution across multiple sources on Bedrock/LangChain, cutting logical-name creation time by 90%.
• Led the enterprise implementation of a RAG platform (SageMaker, LangChain, OpenSearch) across SharePoint and Collibra, accelerating policy Q&A for 50+ compliance analysts.
• Hardened all GenAI pipelines using Model Context Protocol (MCP) for secure context management, alongside CI/CD (GitHub Actions/Jenkins), Infrastructure as Code (Terraform), and AWS Secrets Manager; standardized model/prompt versioning with MLflow/ZenML.
• Implemented event-driven Lambdas to ingest system-catalog deltas, automating data validation and enabling the Data Management Team (DMT) to approve submissions 20% faster.
• Enabled self-service for non-Collibra users by exporting/remodeling spreadsheets and notifying the DMT queue via scheduled jobs, shortening change-detection cycles by several hours.
Environment: SageMaker, Bedrock, LangChain, OpenSearch, Step Functions, Lambda, GitHub Actions, Jenkins, Terraform, Secrets Manager, Python, FastAPIMore...