Professional Experience

• Designed and prototyped a multi-task convolutional neural network (CNN) architecture, building a robust data pipeline to train and integrate the CNN into large-scale linear optimization models, reduced solve time from over 10 to under 2 minutes through CNN-driven approximations of initial solve points.

• Developed an intuitive user interface enabling users to generate and manage input files for distinct simulation models, seamlessly integrating them into a broader modeling ecosystem.

• Enhanced model accuracy by extracting and translating technical documentation into actionable linear optimization constraints, utilizing Python for data preprocessing and GAMS for constraint formulation.

• Constructed a Tableau dashboard for supervisors to monitor on-station officials' activity at terminal stations through integrating data from the MBTA's digital train sheet application. Prototyped a feature which captured user behavior, allowing officials to compare themselves against peers.

• Developed a Python script to batch-download historical prediction files from AWS S3 buckets for Exploratory Data Analysis of MBTA prediction data. Leveraged Matplotlib to visualize prediction accuracy and improve the reliability of information displayed at stations to help combat slow zones.

• Created a service to validate message accuracy on Variable Message Signs (VMS) during the Sumner Tunnel closure, interfaced with MBTA's V3 API and vendor-side logs through Pandas and PostgreSQL.

Lead programming workshops for Python and Rust, aiding students in understanding how to write efficient and effective code and providing guidance on best practices for programming in a data science context.