Dominic Dufour

I build production-grade AI systems that automate high-impact workflows, from “LangGraph-based agent” contract validation to RAG backends that support daily operations for 500+ field technicians. I focus on reliability—using state management, conditional routing, and an evaluation pipeline to measure and improve agent performance.

I design and deliver end-to-end solutions across data ingestion, embeddings, and semantic search, turning unstructured inputs into vector-powered experiences for LLM applications. When needed, I combine ML forecasting with a “text-to-SQL agent” to help users query historical data through natural language and improve estimation accuracy from 50% to 80%.

On the engineering side, I lead MLOps on Azure with architectures on “Azure Databricks and Jenkins,” automating the ML model lifecycle (build, unit testing, deployment, monitoring, periodic retraining). I also serve as the technical lead for a data platform spanning ~10 data sources and 500+ tables, staying hands-on by developing pipelines daily and advising on AI/ML design.

Before transitioning deeper into machine learning, I developed analytics and audit automation by extracting and transforming data from “Salesforce and SAP,” building “Power BI dashboards,” and creating PySpark scripts for duplicate payment detection. That audit background strengthens how I validate assumptions, define risk-aware logic, and evaluate outcomes.