Strawberry Fields

Julian Gao is a machine learning and robotics engineer working where data-centric AI meets autonomous systems. At XPeng he builds the data side of vision-language-model driving: curation pipelines and quality gates that decide what a model learns from, embedding retrieval that traces failures back to the decisions that caused them, and tooling that tells a data problem apart from a model capacity one.

The other half of his work is the system that data feeds. He has spent years on motion planning and control — trajectory optimization, collision avoidance, and lane-centering shipped to production vehicles — and earlier built reinforcement and imitation learning frameworks for robot manipulation, including platforms for collecting human demonstrations at scale. He is now moving towards end-to-end methods for these problems.

Guided by a philosophy of constant growth, he has a mindset as always stay “under training”, believes in living “analog” in a digital world, and is motivated by a personal goal to become an athlete.