Matthew Douglas Green

Developed and owned a telemetry data pipeline to process large, complex telemetry datasets (>100 TB) and implemented a convolutional neural network (CNN) to predict vehicle engine failures with 80% true positive accuracy, now in production to deliver predictive safety insights. Integrated large language models (LLMs; GPT, Llama) with existing ML models (PySpark; SVM, XGBoost) to assess their effectiveness in predicting vehicle safety hazards from text data in customer complaints and warranty claims, resulting in improved overall model performance.

Owned and deployed an end-to-end crop recommendation system using historical and projected weather, soil, and environmental data, translating modeling outputs into actionable guidance for farmers. Built predictive models achieving ~65% accuracy on historical yield prediction and delivered a user-facing application with integrated LLM-powered explanations and natural-language Q&A.

Partnered with product and business teams to identify EV charging failure modes using clustering and distributed analytics, informing decisions projected to drive ~$15M in revenue. Designed and owned an automated demand forecasting pipeline supporting ~$1B in annual revenue, improving forecast accuracy by >40% and enabling ~$10M in annual cost savings.