Machine learning · Streamlit
Youth mental health risk
Scores the chance a youth age 12–17 has a depression, anxiety, or behavior-problem diagnosis, using the NSCH 2024 survey. An XGBoost model explains the score with a SHAP waterfall.
Projects
Six products built to practice applied AI: a youth mental health model, public assistance, a county data agent, a personal-finance lakehouse, calm places in Charlotte, and a household habit tracker. Try the live app, or open README.md and architecture.jpg for each project.
Selected work
Try them
Public funds · Vertex AI
CivicPocket
Helps North Carolina residents match grants and assistance they may already qualify for. Three local questions, a program directory, and an optional agent for a more specific situation.
Agentic RAG · LangGraph
RE Enterprise Architecture RAG
Chat over Mecklenburg County open tax and sales data and North Carolina real-estate license notes. A ReAct agent picks retrieval or SQL instead of stuffing every row into a vector index.
Lakehouse · AWS · Streamlit
FinanceBuddy
Synthetic personal-finance lakehouse: generate ledgers, transform with dbt on Athena, score habits and fraud, and explore gold marts in Streamlit. Optional Kinesis streaming and a Bedrock Haiku coach that only cites numbers already in the marts.
Charlotte · Gemini
Peaceful Places
Non-medical directory of parks, libraries, and community events in Charlotte. Search by zip, filter by noise and energy, and request a Gemini summary of place notes.
On-device · Expo
Rootloop
Habit tracker for one household. Local profiles, daily check-in, streaks and a color report, and sharing by JSON file or QR. No cloud account.