Software Engineer building ML-driven tooling for industrial real estate at Harbor Capital, and CS student at Duke University (concentration in AI/ML).
- Underwriting automation — LightGBM/Optuna forecasting models, LangGraph multi-agent pipelines for offering-memo extraction, and a pgvector semantic search engine over 18K+ asset records
- Internal tooling — a Kubernetes operator (Go, kubebuilder) that turns dropped files into auto-retrying document pipelines
- Product @ Duke (Operations Director)
Python · TypeScript · C · SQL · React / Next.js · FastAPI · PyTorch · PostgreSQL · AWS · Docker · LangGraph / RAG
| Project | What it does |
|---|---|
| Harbor-Capital-Scraper | Streamlit tool that ingests, geocodes, and maps broker comp spreadsheets into a searchable Supabase comp database |
| Legends_NPoints | Full-stack pipeline (FastAPI + React) clustering unstructured social posts into psychographic micro-personas via BERTopic/UMAP/HDBSCAN |
| HarborCap_ActivityReportTracker | FastAPI + React dashboard for tracking deal/broker activity reports, deployed on Railway |
| HarborCap_Bullpen | Multi-agent CRE intelligence platform — Python agents + dashboard for deal sourcing and market feed |
| dumpingdesk | MVP automating U.S. Dept. of Commerce antidumping questionnaire workflows (Next.js + FastAPI) |
| PawsitiveProduct | Product @ Duke's pet-care MVP (Next.js, Supabase) |