STATE OF MISSOURI ITSD • AI DEVELOPMENT • MACHINE LEARNING
I am a Business Analyst with Missouri's ITSD, currently supporting applications for the Department of Elementary and Secondary Education. Before that I built XGBoost models for billion-dollar mortgage portfolios at Veterans United, ran a coffee business for 12 years, worked as an electrician, and studied economics. The common thread is that I like taking complicated systems apart and figuring out how to make them work better.
My current focus is AI and legacy system modernization. I recently refactored part of the School Finance application's data pipeline, replacing record-level Visual Basic calculations routed through 6 SSIS packages with array-level SQL, reducing database calls from over 300,000 to a handful. On my own time I built a production-ready RAG system that processes 22.6 million words of Missouri education policy and statute documentation, with GPU-accelerated embeddings, hybrid search, and citation enforcement. It runs locally on Linux and is designed for FERPA compliance.
I hold an MA in Economics from UMSL with a 3.914 GPA and graduate certificates in AI, Data Science, and Applied Econometrics.
MA Economics
UMSL • 3.914 GPA
3 Graduate Certificates
Business Analyst
Missouri ITSD
DESE Applications
Microsoft AZ-900
Azure Fundamentals
gbmframework
Published on PyPI
8B+ Observations
GNMA Mortgage Data
75% Load Time Reduction
$1M Annual Revenue
12 Years Ownership
20 Employees
Python
SQL
R
XGBoost
SHAP
PyTorch
ChromaDB
sentence-transformers
Ollama
CUDA
Tableau
Power BI
Snowflake
PySpark
ArcGIS Pro
Git
MongoDB
EViews
Excel
Background
Projects
Business Analyst
ITSD, Office of Administration (Aug 2025 – Present)
Supporting DESE applications. Refactored School Finance data pipeline, reducing 300,000+ database calls to a handful. Identified and remediated PII security vulnerabilities in production systems. Built RAG system for education policy on personal time.
Capital Markets Analyst
Veterans United Home Loans (Jul 2024 – Feb 2025)
Built XGBoost delinquency prediction models for billion-dollar mortgage portfolios. Optimized GNMA dashboard from 8-hour to 2-hour load time on 8B+ observations. Created Python automation saving 7–9 hours weekly.
MA Economics – University of Missouri-St. Louis (2024)
Certificates: Data Science, Applied Econometrics, Artificial Intelligence
BA Economics – University of Cincinnati
Doctoral coursework – University of Illinois Chicago
AI/ML: RAG systems, LLM integration, XGBoost, gradient boosting, SHAP explainability
Data: SQL optimization for billion-scale datasets, ETL pipeline development, SSIS
Methods: Causal inference, econometrics, time series, A/B testing
Production-ready retrieval-augmented generation system processing 22.6 million words of Missouri education statutes and policy documents. GPU-accelerated embeddings with sentence-transformers, hybrid search combining vector similarity with keyword boosting for statute number detection, and citation enforcement so every answer traces to its source. Runs locally on Linux with an 8GB GPU, designed for FERPA compliance.
Python
RAG
ChromaDB
sentence-transformers
PyTorch
CUDA
Open-source Python package unifying XGBoost, LightGBM, and CatBoost with automated hyperparameter optimization, SHAP analysis, and system resource detection. Published on PyPI.
Python
Machine Learning
PyPI
Interactive Tableau dashboard analyzing government-backed mortgage portfolio performance. Optimized SQL queries for 8B+ observations, reducing load time by 75%.
Tableau
SQL
Financial Analysis
Geospatial analysis combining ArcGIS Pro and Python web scraping to optimize business location using demographic data and market gap analysis.
ArcGIS Pro
Python
Geospatial
Comprehensive benefit-cost analysis of municipal fleet transition from ICE to electric vehicles for the City of Kirkwood.
Python
Cost-Benefit Analysis
Policy
Performance evaluation of gradient boosting algorithms across multiple datasets using the GBM Framework.
Machine Learning
Benchmarking
Python
© 2025 Mark Attwood