• Supported a graduate machine learning course by helping students build, debug, and evaluate models in Python using real-world datasets.

• Provided detailed technical feedback on assignments and projects involving feature engineering, model training, and performance evaluation.

Built a probabilistic forecasting pipeline combining historical, weather, and LLM-extracted news signals for outcome prediction.

Developed a FastAPI inference service that converted real-time news into structured features, improving calibration by 12%.

Built XGBoost weather-impact models across temperature, humidity, and precipitation, reducing scenario variance by 17%.

Added a Bayesian output layer to model uncertainty, improving calibration by 15% and reaching a Brier score of 0.176 with 67%

accuracy.

Improved dominant-outcome recall from 82% to 84% while reducing overconfident predictions in uncertain cases.

• Designed a regression-based model to analyze cycle time data, identifying bottlenecks and predicting station delays.

• Transformed unstructured worker logs into structured datasets, enhancing data-driven analysis capabilities.

• Developed standardized data pre-processing pipelines for consistent data collection, improving reliability across teams.

Trained and mentored multiple technician teams by formalizing critical maintenance procedures into standardized, repeatable workflows, ensuring safe, consistent, and low-error execution across mechanical systems.

Planned and managed preventive and corrective maintenance schedules, analyzing failure patterns to minimize downtime and improve overall system reliability.

Redesigned the maintenance shop layout as a process optimization problem, improving workflow efficiency and accessibility, and reducing maintenance lead times while increasing technician productivity by 12%.

Designed and validated custom diagnostic gauges and jigs to improve measurement accuracy and alignment, reducing manual error rates and enabling reliable, high-quality diagnostic signals during tractor servicing.

• Designed and drafted fiber optic network layouts for the Southeast U.S. region using Aramis and AutoCAD

software, ensuring precise routing and optimized infrastructure for RF communication.

• Ranked top in training, completing all assigned design tasks faster than peers with 100% accuracy,

demonstrating strong technical aptitude and rapid learning in telecom infrastructure planning.