Maria Ann Toms

Data Scientist @DNV

Katy, TX, US

Architected multi-agent LLM platform (LangChain + Azure OpenAI) reducing manual configuration by 75% and supporting 500+ users.Developed MCP server with JSON-RPC, OpenAPI validation, and Redis cluster for concurrent agent workflows.Fine-tuned AWS Bedrock models to auto-generate JavaScript business rules, reducing analyst coding from 2 hours to 5 minutes.Built production RAG pipeline vectorizing 50K+ tickets using Azure OpenAI embeddings, stored in MongoDB Atlas Vector Search.Engineered semantic retrieval system achieving 90% context relevance for resolution recommendations.