CommentGPT
Uline | 2024
Uline’s first enterprise AI agent — a fully orchestrated LLM-based system analyzing 2M+ annual customer comments in near-real time. Spans product, service, shipping, and technology feedback channels.
- Reduced time to identification of critical issues by 98% (from two days to one hour)
- Lifted caught issues by over 31% compared to the prior manual process
- Built with Azure OpenAI, LangChain, and custom orchestration logic
Enterprise RAG Architecture
Uline | 2025
Production-scale Retrieval-Augmented Generation system serving 2.5M annual queries against internal customer service procedure knowledge bases.
- Reduced employee time-to-answer by 73%
- Decreased internal helpline escalations with instant, cited answers
- Hybrid search architecture using SQL Server 2025 vector and full text search, combining semantic and keyword retrieval
DMA 2016 Analytics Challenge
Direct Marketing Association | 2016
Competitive analytics challenge — placed 4th out of 50 teams.
- Developed ensemble models for loan profitability prediction
- Dataset of 700,000+ records
- Combined gradient boosting, random forests, and logistic regression
Transportation Optimization — Johnson Controls
Marquette University | 2014
Graduate-level operations research project applying linear programming to a real-world logistics problem.
- Modeled Johnson Controls’ transportation network
- Optimization resulted in an estimated $1M in annual savings
- Built with R and linear programming solvers