I build agentic AI systems that improve learning, support, and operational outcomes. At Heartland Community Network, I engineered a LangGraph AI tutor that increased session completion by 48% and reduced unresolved-query backlog by 30%.
At Indiana University's AI and Natural Language Processing Lab, I designed a Graph-RAG system over 1,000+ PubMed records, reaching 0.84 RAGAS faithfulness. I also trained an XGBoost hypertension model using GatorTron clinical embeddings and PCA-reduced features, achieving 0.90 AUROC through 5-fold stratified cross-validation.
Previously at Deloitte Touche Tomatsu, I built NLP recommendation, PII masking, OCR redaction, and analytics solutions for enterprise support workflows. My production multi-agent system on Vertex AI Agent Engine delivered an estimated $1.3M+ in annual cost savings.
I work across Python, LangChain, LangGraph, FastAPI, Vertex AI, Neo4j, and evaluation-driven ML to turn complex AI capabilities into measurable results.
