Cruise LLC (GM)
Cruise Data Science — System Performance Validation (Full‑time role)
Data pipelines and models for fleet safety, triage, and ride comfort.
About the Project
Autonomous vehicles continue to grow in fleet distribution and environment. Monitoring core metrics for system performance was the main purvuew of the Cruise Data Science team.
The Cruise AI ecosystem involved multiple system components that needed to integrate seamlessly in simulation and in real-world performance. Identifying and reducing the number of "take-over events" was a mission-critical goal. Part of this goal was to identify and classify the perception and actions of the system as a whole, which fell to the Cruise Data Science team.
As a Staff Data Scientist, I led pipeline development to identify ride quality events, evaluate them in test simulations, and have equivalent comparison for on-road data.
I led "special forces" data science teams to help cross-disciplinary projects resolve complex data problems. This included a more accurate alignment of 2D object identification along with lidar/radar distance in annotated data.
Project Execution
• Explored and defined what "Ride Comfort" means to users; created new top-line metrics to maintain ride comfort performance spanning sim/real.
• Feedback loop from ops to model teams.
• Created preliminary take-over event classifier, identifying the type of event requiring driver intervention. Pareto frequency dashboards provided real time information on current system challenges.