About me
Five years across deep learning for medical diagnostics, a graduate teaching assistantship at the University of Cincinnati, and production clinical AI at Zasti — agent pipelines with a human at every consequential action, and multimodal research that holds up under review. Let’s build something incredible together!
What I do
Agentic systems
A shared agent harness — LangGraph through MCP — carrying four automation pipelines at Zasti: sales, resource tracking, marketing, and internal software engineering. Adopted company-wide, with a human approval gate before any consequential action.
Multimodal retrieval
Production RAPTOR RAG with hierarchical summarization, knowledge-graph integration, and dual-stage vector retrieval — 50M+ document embeddings served at sub-100ms P99 latency.
Clinical AI
Multimodal clinical pipelines built on GPT-4V vision and FHIR-compliant parsers, with hallucination detection via chain-of-thought consistency scoring — 92% retrieval precision across 15+ heterogeneous lab-report formats.
Research & evaluation
The 2026 preprints: valid per-field selective risk control for document extraction (arXiv:2608.14639), and an audit of automated evaluation in tool-using agents across 14,750 execution traces (arXiv:2604.16706).
Teaching
Graduate teaching assistant at the University of Cincinnati: hands-on Python, cloud, and ML workshops for 500+ graduate students, and 20+ capstone research projects mentored.
