Yuye (Julia) Zhou

Multimodal Urban Perception for Autonomous Systems (UC Berkeley, Jan 2025– )

• Developed a causal graph fusion framework integrating LiDAR, panoramic imagery, and satellite data.
• Designed joint evaluation protocols for dynamic robustness and static scene understanding (SVF, occlusion).
• Produced interpretable modality attributions to enhance safety-critical perception reliability.

Large-Scale LiDAR Processing for AV Line-of-Sight Analysis (UC Berkeley, Oct 2024– )

• Built scalable Python pipelines to process 900+ LiDAR (LAZ) files in San Francisco, extended to 10+ U.S. cities.
• Computed Sky View Factor and quantified occlusions from urban morphology for AV sensor placement.
• Produced high-resolution DSM/DEMs for line-of-sight and coverage validation.

Autonomous Vehicle Collision Data Analysis (2025– )

• Analyzed California DMV AV crash reports and geospatial infrastructure data.
• Engineered features from road geometry, visibility, and infrastructure to identify high-risk zones.
• Developed insights for safer AV deployment and urban design improvements.

Multimodal AI-Agent Framework for Urban Simulation (UC Berkeley, Feb 2025– )

• Designed an AI-agent framework to simulate human-environment interactions at city scale.
• Fused LiDAR morphology, street-view imagery, and satellite data for agent perception.
• Integrated LLM-driven behavioral profiles to create diverse, realistic simulations.

Green Schoolyard Equity & Shade Access (UC Berkeley, 2024– )

• Mapped and analyzed shade equity across New Mexico schoolyards using GIS and remote sensing.
• Integrated tree canopy, built environment, and demographic data to assess exposure disparities.
• Collaborated with Green Schoolyards America to inform equitable design interventions.