Predicting User Preferences with Deep Learning from Vehicle Telemetry Data

Modern vehicles are becoming sensor-rich data platforms, creating an opportunity to move beyond reactive, manual user controls. During my 8-month Data Scientist co-op term at Rivian & Volkswagen Tech (RVT), my primary project focused on exploring Proactive Personalization by developing a deep learning approach to predict in-cabin user preferences using historical interaction patterns and real-time contextual signals.

My primary contribution was the design and implementation of an end-to-end machine learning pipeline. This included defining the problem, constructing a large-scale ETL workflow for terabyte-level telemetry data using PySpark and SQL, developing a neural network model in TensorFlow/Keras, and iteratively refining the system through evaluation and validation. I also built dashboarding tools and presented findings across teams to support cross-functional understanding.

To join this seminar virtually, please request Zoom connection details from ea@stat.ubc.ca.