At Northstar Ingredient Intelligence, I defined a reference architecture for retrieval-augmented agents combining Azure OpenAI, Azure AI Search, PostgreSQL, enterprise APIs, and human approval checkpoints. I reworked supplier-document retrieval, raising grounded-answer acceptance from 68% to 86% across a controlled evaluation set of compliance questions.
I built reusable Python services and deployment templates for connecting agents to enterprise APIs, with shared security and observability. By combining extraction, evidence-linked summaries, deterministic policy checks, and human sign-off, I reduced analyst review time for ingredient-change workflows from 42 minutes to 17 minutes.
Before Northstar, I designed a multi-tenant document intelligence platform at Harborline Commerce Systems and built data services at Cobalt Data Works, Meridian Integration Group, and Pioneer Digital Services. Across these roles, I improved workflow reliability and performance, developed production integrations, and introduced testing and evaluation practices; I also hold Microsoft Azure AI and Solutions Architect certifications and an AWS Machine Learning certification.

