At MidstreamAI, I built and owned a production-grade multi-agent analytics platform that lets oil and gas operators query pipeline infrastructure and telemetry in natural language across approximately 30 domain APIs and tools. I improved LLM tool-selection accuracy from 55% to 95%, reduced query latency to 5–12 seconds, and cut token consumption by approximately 60%.
I also build evidence-grounded RAG systems and secure enterprise applications with Python, FastAPI, PostgreSQL, LangGraph, React, and ASP.NET Core. My work spans pipeline monitoring, industrial regulatory knowledge retrieval, and AI-powered employee wellness workflows.
