10+ years of strategy and operations. An MBA and an MSAI in progress at KSU. I don't start with models — I start with the business problem. Every pipeline I build ends with a dollar figure or an executive-ready recommendation.
I'm not a typical data scientist. Before my MSAI at Kennesaw State University, I spent over a decade in the field — first as a statistician at a national railway authority in southern Africa, where I rebuilt the company's pricing model from scratch and earned the CEO's Employee of the Year Award. Then as a General Manager who built demand forecasting systems that drove 15–20% weekly sales growth.
That background shapes everything I build. I don't start with models — I start with the question: what decision does this data need to enable? Every pipeline I write ends with a dollar figure, an operational recommendation, or an executive-ready dashboard.
I believe the most dangerous data scientist is one who can't explain their model to a room full of people who don't care about MAPE. The most valuable one is someone who can do both: build the system and tell the story.
Currently seeking Data Scientist and AI Strategy roles across Georgia, Atlanta Metro, and Remote.
Every project starts with a real business problem. Every result is measured in dollars, not just metrics.
End-to-end ML pipeline — raw data ingestion, feature engineering (19 features including lag, rolling, calendar), Prophet + XGBoost ensemble, MLflow experiment tracking, and a Streamlit dashboard with business dollar-impact summary. MAPE under 10%.
3-block CNN with Grad-CAM explainability — surfaces deep feature activations as visual diagnostic heatmaps. Enables non-technical stakeholders to validate model behaviour. Deployed as a containerised REST API via FastAPI and Docker.
Production-grade ML model served as a REST API — containerised with Docker, automatically deployed via GitHub Actions CI/CD to Google Cloud Run. Load-balanced routing, health checks, and automated rollback on failure.
Regression-based pricing model that translates raw cost and demand data into an executive-facing revenue strategy. Includes an interactive what-if scenario tool for operations teams. Inspired by real work at a national railway authority in southern Africa.
Forecasting pipeline scaled to distributed compute using PySpark. Delta Lake data management, MLflow experiment tracking, and scalable feature engineering across large multi-store retail datasets. Runs on Databricks Community Edition.
Retrieval-Augmented Generation system that lets business users ask plain English questions about operational data and receive grounded, cited answers — powered by LangChain, ChromaDB, and Sentence Transformers. No SQL required. No dashboard needed. Production deployment guide included for Claude and GPT-4 integration.
LLM-powered agent that takes ML model outputs and SHAP feature attributions and automatically generates plain English stakeholder briefings — translating every prediction into a narrative that a non-technical decision-maker can read and act on immediately. Includes SHAP visualisation dashboard and production Claude/GPT-4 integration patterns.
Built over 10+ years across statistics, operations, and AI engineering. Every skill below is demonstrated in a live project.
I write about the intersection of data science, business strategy, and operational leadership.
Open to Data Scientist and AI Strategy roles in Georgia · Atlanta Metro · Remote. Happy to talk about your data problems before we talk about job titles.
Send me a message →