MIS student at the University of Georgia, Terry College of Business (class of 2027), interested in AI systems and evaluation, risk and insurance analytics, IT audit and GRC, and BI/analyst work.
Each project below ships with tests, the commands to reproduce its results, and a limitations section. Where data is synthetic or simulated, the README says so up front.
| Lane | Project | What it is |
|---|---|---|
| Risk & insurance analytics | auto-pricing-model-risk | Poisson/Gamma GLM vs LightGBM pricing on 669k real motor policies, with a model-risk validation and a governance pack (NAIC, NIST AI RMF). |
| Risk & insurance analytics | captive-feasibility-model | Monte Carlo comparison of guaranteed cost, large deductible and captive for a hypothetical $50M manufacturer (simulated losses), with Excel workbook and CFO memo. |
| Risk & insurance analytics | cyber-risk-quant | FAIR-style Monte Carlo that converts cyber-insurance application controls into dollar loss estimates and a ranked fix list (hypothetical firms). |
| IT audit & security | itgc-audit-lab | ITGC audit of a lab I built (Samba AD, PostgreSQL ERP, git/CI): live evidence collection, sampling, workpapers and management letter. Synthetic company. |
| AI & LLM evaluation | llm-redteam-risk-register | Red-team of a tool-using LLM assistant (60 attacks, 5 control layers), mapped to OWASP LLM Top 10, MITRE ATLAS and NIST AI RMF. Synthetic target. |
| AI & LLM evaluation | 10k-risk-radar | LLM pipeline over 176 SEC 10-Ks that classifies risk factors and tracks year-over-year changes, scored against a hand-labeled gold set. |
| AI & LLM evaluation | insurance-policy-qa | Clause-cited RAG over real FEMA flood policies that refuses when the documents are silent, measured on a 120-question evaluation set. |
| Data engineering & BI | nfip-flood-claims-analytics | dbt + DuckDB star schema over 2.7M FEMA flood claims with loss trends, a severity model, a dashboard and a Power BI-ready export. |
Python, SQL (PostgreSQL, DuckDB), dbt, Excel, Power BI, Streamlit, LightGBM, SHAP, Docker, PowerShell, pytest, GitHub Actions, Ollama (local LLMs).